Limited-Time Offer: Save up to 55% and Enjoy Seedance from Just $0.02/sec.Try Now

SeeGen AI API

Welcome to the SeeGen AI API — one API for cinematic video and high-fidelity image generation across multiple leading models.

Overview

SeeGen AI is a unified generation API: one consistent set of endpoints to run multiple leading AI video and image models. Choose a model with the model parameter — authentication, task submission, status polling, and webhooks work the same way across all of them.

Currently available models:

  • Video — Seedance 2.5 (sd2.5), Seedance 2.0 Pro (sd2), Fast (sd2-fast), Mini (sd2-mini), Grok Imagine Video 1.5 (grok-imagine-video-1.5) and Lite (grok-imagine-video-1.5-lite), Wan 3.0 Video Prime (wan3.0-video-prime), Wan 3.0 Video (wan3.0-video). Grok supports 1–15 seconds with audio. Standard 1.5 supports native 1080p for text or single-image input and up to 14 reference images; Lite supports text or one image, with 1080p upscaled from 720p.
  • Image — GPT Image 2 (gpt-image-2), Nano Banana 2.1 / 2 (nano-banana-2.1, nano-banana-2), Nano Banana Pro (nano-banana-pro), Seedream 5.0 Lite (seedream-v5.0-lite), Seedream 5.0 Pro (seedream-v5.0-pro) Qwen Image 2.1 Pro and Turbo support 1K/2K generation and editing with up to 10 reference images.

Base URL: https://seegen.ai/api/v1

Model: pass any alias above as the model field (e.g. sd2, gpt-image-2).

Why SeeGen AI?

More Reasons to Choose SeeGen AI:

  • More than Seedance 2.0: Wan 3.0, Nano Banana, ChatGPT Image, and more.
  • No subscription is required — pay as you go
  • Fast access to the latest models
  • 24/7 customer support
  • Support for official model-related inquiries
  • Console access for developers
  • Open to both businesses and individual users

Pricing

40% OFF

API Pack

$500$833

125,000 credits ($0.004/credit)

~1,136 5s videos

40% OFF

API-XL Pack

$2,000$3,332

500,000 credits ($0.004/credit)

~4,545 5s videos

Seedance 2.0 / Fast / Mini / 2.5 — cost calculation

Let out = output seconds and in = the sum of each input video duration rounded up to a whole second, with a minimum of ⌈2 × out / 3⌉. Without video input, cost = out × rate; with video input, cost = (out + in) × rate. Use the model and resolution rates below.

No video inputInclude video input
sd2.5-flash: 480p20 × out15 × (out + in)
sd2.5-flash: 720p40 × out30 × (out + in)
sd2.5: 480P30 × out23 × (out + in)
sd2.5: 720P60 × out45 × (out + in)
sd2.5: 1080P (native)150 × out113 × (out + in)
sd2.5: 2K(60 + 30) × out45 × (out + in) + 30 × out
sd2.5: 4K (native)241 × out187 × (out + in)
sd2-pro: 480P20 × out15 × (out + in)
sd2-pro: 720P40 × out30 × (out + in)
sd2-pro: 1080P (native)100 × out75 × (out + in)
sd2-pro: 2K(40 + 30) × out30 × (out + in) + 30 × out
sd2-pro: 4K (native)200 × out150 × (out + in)
sd2-fast: 480P11 × out16 × out8.5 × (out + in)12 × (out + in)
sd2-fast: 720P22 × out32 × out17 × (out + in)24 × (out + in)
sd2-fast: 1080P(22 + 20) × out(32 + 20) × out17 × (out + in) + 20 × out24 × (out + in) + 20 × out
sd2-fast: 2K(22 + 30) × out(32 + 30) × out17 × (out + in) + 30 × out24 × (out + in) + 30 × out
sd2-fast: 4K(22 + 40) × out(32 + 40) × out17 × (out + in) + 40 × out24 × (out + in) + 40 × out
sd2-mini: 480P4.5 × out10 × out3.5 × (out + in)7.5 × (out + in)
sd2-mini: 720P9 × out20 × out7 × (out + in)15 × (out + in)
sd2-mini: 1080P(9 + 20) × out(20 + 20) × out7 × (out + in) + 20 × out15 × (out + in) + 20 × out
sd2-mini: 2K(9 + 30) × out(20 + 30) × out7 × (out + in) + 30 × out15 × (out + in) + 30 × out
sd2-mini: 4K(9 + 40) × out(20 + 40) × out7 × (out + in) + 40 × out15 × (out + in) + 40 × out

Note: sd2 and sd2.5 generate native 1080P and 4K. Their 2K output and all sd2-fast / sd2-mini 1080P/2K/4K output use automatic upscale from 720P. Choose any tier with outputResolution, for example "4k". Native tiers have no upscale fee. Upscaled tiers add +20 / +30 / +40 credits per output second for 1080P / 2K / 4K. If automatic upscale fails, the completed task returns the 720P fallback without a partial credit refund.

Limited-time offer: sd2-fast and sd2-mini base rates are temporarily reduced. Without video input, Fast costs 11 / 22 credits per output second at 480P / 720P, and Mini costs 4.5 / 9 credits per output second. With video input, Fast costs 8.5 / 17 credits per billed second, and Mini costs 3.5 / 7 credits per billed second, applied to the combined output and billable input-video duration. For Fast and Mini, 1080P/2K/4K use the discounted 720P base rate plus an unchanged upscale add-on of 20 / 30 / 40 credits per output second.

If upscaling fails, the completed task returns the 720P fallback without a partial credit refund.

  • sd2.5 480P, 4-second output without video input: 120 credits
  • sd2.5 720P, 5-second output without video input: 300 credits
  • sd2.5 720P, 5-second output + 3-second input video (4-second minimum input): 405 credits
  • sd2.5 1080P native, 5-second output without video input: 750 credits
  • sd2.5 1080P native, 5-second output + 5-second input video: 1,130 credits

MiniMax H3 / H3 Max / H3 Max Turbo

MiniMax H3
ResolutionOutput credits / secondReference video credits / second
768P2727
2K4343

First 5 images free; 11 credits per additional image.

MiniMax H3 Max
ResolutionOutput credits / secondReference video credits / second
480P1820
768P2752

First 2 images free; 27 credits per additional image.

MiniMax H3 Max Turbo
ResolutionOutput credits / secondReference video credits / second
480P10—
768P14—

First 2 images free; 27 credits per additional image.

Credits = output seconds × output rate + sum of each input video’s duration rounded up to a whole second × input video rate + images beyond the free allowance × image rate. All input images count, including first/last frames. Audio references are free. Server-measured video duration determines the charged amount; client-supplied durations are ignored. Failed tasks refund the reserved credits.

Kling 4.0

Credits / sec
ResolutionNo video inputInclude video input
720P5432
1080P8048

Without input video: 720p 54 and 1080p 80 credits per output second. With input video: 720p 32 and 1080p 48 credits per billable second. The server measures each input video and rounds its duration up to a whole second before summing. Billable seconds = output + max(rounded input total, output). Final credits are rounded up. The 30-second input limit uses actual durations before rounding.

Try Kling 4.0 in the API Playground

Grok Imagine Video 1.5

ResolutionCredits / sec
480P30
720P50
1080P90

Output costs 30 / 50 / 90 credits per second at 480p / 720p / 1080p, plus 5 credits per input image. Audio does not change the price. A 5-second 480p video with one input image costs 155 credits.

urls: public HTTP(S) image URLs. One image selects single-image-to-video; two images select first/last-frame. Set videoInputMode: "reference" for 1–14 reference images. First/last-frame and multi-reference output support 480p and 720p.

Grok Imagine Video 1.5 Lite

ResolutionCredits / sec
480P8
720P12
1080P50

Output costs 8 / 12 / 50 credits per second at 480p / 720p / 1080p, plus 5 credits per input image. Audio does not change the price. A 5-second 480p video with one input image costs 45 credits.

urls: zero or one public HTTP(S) image URL. One image selects image-to-video and follows its aspect ratio. Multiple images, reference mode, video and audio inputs are not supported.

Wan 3.0 Video / Prime — cost calculation

Let out = output seconds and in = the total reference-video seconds, with each clip rounded up to a whole second. Without reference video: credits = out × rate. With reference video: credits = (out + in) × rate. Use the rate for your model and resolution below.

ModelNative outputCredits / sec2s example
wan3.0-video-prime480P2244
720P4590
1080P90180
wan3.0-video480P1632
720P3264
1080P64128

Note: 480P, 720P, and 1080P are native output; 2K/4K upscale is not supported. Reference images and audio do not add to the billed duration. Turning generated audio off does not change the price. Reference-video duration plus output duration must not exceed 30 seconds.

  • wan3.0-video 480P, 2-second output without video input: 2 × 16 = 32 credits
  • wan3.0-video-prime 480P, 2-second output without video input: 2 × 22 = 44 credits
  • wan3.0-video 720P, 5-second output + 3-second reference video: (5 + 3) × 32 = 256 credits
  • wan3.0-video-prime 720P, 5-second output + 2.2-second reference video (rounded up to 3 seconds): (5 + 3) × 45 = 360 credits

🎉 Limited-time offer: 33% off all image generation — all image prices below are already discounted.

GPT Image 2.5 Flare / Sunburst

ResolutionMedium qualityHigh qualityXHigh qualityMax quality
1k35132023355075
2k (default)7102335375584125
4k1015406067100154230

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically.

gpt-image-2 — credits per image

ResolutionMedium qualityHigh quality
1k10154770
2k (default)233584125
4k4060154230

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically.

Nano Banana — credits per image

Resolutionnano-banana-2.1nano-banana-2nano-banana-pro
1k20304060
2k (default)30454060
4k477074110

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically.

Seedream 5.5 — credits per image

Item / tierCredits
1k1015
2k (default)1725
Reference image (each)46

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically. All reference images are billed, including the first.

Seedream 5.0 — credits per image

Model / tierCredits
Flash 1k / 1.5k / 2k (flat rate)812
Lite 2k / 4k (flat rate)710
Pro 1k (default)1015
Pro 2k2030

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically. Reference images are not billed separately.

Qwen Image 2.1 Pro / Turbo

Resolutionqwen-image-2.1-proqwen-image-2.1-turbo
1k132058
2k132058

Pro: 13 credits per image. Turbo: 5 credits per image. 1K and 2K cost the same. Prices include any active image promotion.

Each task produces 1 image. Submit N tasks for N variants. Failed tasks refund credits automatically. Reference images are not billed separately.

Check your balance: GET /api/v1/account/credits

Automatic Recharge for API Accounts

Your API key and SeeGen AI dashboard use the same account-level credit balance. Auto Recharge can keep that balance funded for API workloads, but it must be configured from the Credits page; there is currently no separate configuration API.

Set up Auto Recharge from Credits:

  1. Open Credits, choose the minimum balance you want to maintain, and select a recharge package.
  2. Authorize your payment method once. This setup step saves authorization and does not charge the card.
  3. After the flexible authorization is active, you can change the minimum balance or package from Credits without authorizing again. An account using the previous fixed-price authorization may need a one-time upgrade.

How it behaves with API requests

  • When a successful API task submission deducts credits and moves the balance from at or above your threshold to below it, SeeGen AI queues an automatic recharge asynchronously. The task submission does not wait for the recharge.
  • If the account does not have enough credits before submission, the API returns HTTP 402 and creates no task. Auto Recharge is not triggered by that rejected request, and the request is not automatically retried; retry it after credits are available.
  • Check the current balance with GET /api/v1/account/credits. Automatic recharge progress and failures appear on Credits and Payment History, and important outcomes are emailed to the account billing address. There is currently no customer-facing automatic recharge webhook or status API.

Automatic recharge bonuses for API packages

  • $500.00 API Pack: 127,500 credits (125,000 + 2% bonus).
  • $2,000.00 API-XL Pack: 525,000 credits (500,000 + 5% bonus).
  • Manual purchases of these packages still add 125,000 and 500,000 credits respectively.

Authentication

All API requests require a Bearer token in the Authorization header. You can create and manage API keys from your Account Settings.

curl -H "Authorization: Bearer YOUR_API_KEY" \
  https://seegen.ai/api/v1/account/credits

Important: Your API key is shown only once at creation. Store it securely. You can create up to 10 API keys per account.

Quick Start

Generate a video in two steps: create a task, then poll for the result.

set -eu
# 1. Create once. Save the returned IDs; never blindly retry a timed-out create.
CREATED=$(curl --fail-with-body --silent --show-error --max-time 120 -X POST \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "inputs": {
      "prompt": "A golden retriever running on the beach at sunset",
      "duration": "5s",
      "resolution": "1280x720"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask)
printf '%s\n' "$CREATED" | jq .
TASK_ID=$(printf '%s' "$CREATED" | jq -er '.taskId')

# 2. Poll for up to 15 minutes. A polling timeout does not cancel the task.
DEADLINE=$(( $(date +%s) + 900 ))
while true; do
  RESULT=$(curl --fail-with-body --silent --show-error --max-time 30 \
    -H "Authorization: Bearer $SEEGEN_API_KEY" \
    "https://seegen.ai/api/v1/jobs/queryTask?taskId=$TASK_ID")
  STATUS=$(printf '%s' "$RESULT" | jq -er '.status')
  if [ "$STATUS" = "COMPLETED" ]; then
    printf '%s\n' "$RESULT" | jq .
    break
  fi
  if [ "$STATUS" = "FAILED" ]; then
    printf '%s\n' "$RESULT" | jq .
    exit 1
  fi
  if [ "$(date +%s)" -ge "$DEADLINE" ]; then
    echo "Still processing: query taskId=$TASK_ID later" >&2
    exit 1
  fi
  sleep 5
done

Endpoints

POST/api/v1/jobs/createTask

Create a new image or video generation task

GET/api/v1/jobs/queryTask

Query task status and result

GET/api/v1/jobs/relations

Query draft and final task relations

GET/api/v1/account/credits

Check your credits balance

POST/api/v1/assets/upload

Upload an asset (image/video/audio) for review

GET/api/v1/assets/status

Query asset review status

POST/api/v1/assets/review

Start or retry platform material review

GET/api/v1/assets/list

List your uploaded assets

POST/api/v1/upscale/create

Submit a standalone video upscale task (720p / 1080p / 2K / 4K)

GET/api/v1/upscale/query

Poll a standalone upscale task's status and result

Choose a Model (Video)

Choose a video model by workflow, native resolution, speed, and price; combine this table with the Pricing section to estimate cost.

ModelT2VI2VFirst–LastMulti-RefR2VNative 1080pNative 4KAudio720p / 5s
sd2.5✓✓✓✓—✓✓✓300 credits
sd2.5-flash✓✓✓✓———✓200 credits
sd2✓✓✓✓—✓✓✓200 credits
sd2-fast✓✓✓✓—upscaleupscale✓110 credits160 credits
sd2-mini✓✓✓✓—upscaleupscale✓45 credits100 credits
grok-imagine-video-1.5✓✓✓✓—✓—✓250 credits
grok-imagine-video-1.5-lite✓✓———upscale—✓60 credits
wan3.0-video-prime✓✓✓✓—✓—✓225 credits
wan3.0-video✓✓✓✓—✓—✓160 credits

T2V = Text to Video · I2V = Image to Video (first frame) · First–Last = first + last keyframe · Multi-Ref = reference images (Seedance and Wan also support video / audio) · R2V = 1–9 reference images with character markers

Seedance 2.0 / 2.0 Fast / 2.0 Mini / Seedance 2.5

Text to Video

Generate a video from a text prompt. No images required.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "inputs": {
      "prompt": "A futuristic city with flying cars at night, neon lights reflecting on wet streets",
      "duration": "5s",
      "resolution": "1280x720"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask
ParameterTypeRequiredDescription
promptstringYesText description of the video to generate. Max 20000 chars.
durationstringNoVideo length: "4s" to "15s"; sd2.5 supports up to "30s" (default: "5s")
resolutionstringNoAspect ratio via resolution. Options: auto (default), 720x720, 720x960, 960x720, 1280x720, 720x1280, 1280x540
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.
seedintNoReproducibility seed: -1 or omit for random; 0–2147483647 for fixed.
generateAudiobooleanNoWhether to synthesize a synced audio track. Default true; pass false for silent video.

Image to Video

Animate a static image into a video. Provide one image URL as the starting frame.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "inputs": {
      "urls": ["https://example.com/photo.jpg"],
      "prompt": "The woman slowly turns her head and smiles",
      "duration": "5s"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask
ParameterTypeRequiredDescription
urlsstring[]YesArray with one image URL (the source frame). Supports both HTTP URLs and asset references (e.g. "asset://asset-20260326-abc123")
promptstringNoText description of the desired motion
durationstringNoVideo length: "4s" to "15s"; sd2.5 supports up to "30s" (default: "5s")
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.
seedintNoReproducibility seed: -1 or omit for random; 0–2147483647 for fixed.
generateAudiobooleanNoWhether to synthesize a synced audio track. Default true; pass false for silent video.

First & Last Frame

Define the starting and ending frames, and the model generates the transition between them. Uses videoInputMode: "keyframe".

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "inputs": {
      "urls": [
        "https://example.com/first-frame.jpg",
        "https://example.com/last-frame.jpg"
      ],
      "prompt": "Smooth camera transition from day to night",
      "duration": "5s",
      "videoInputMode": "keyframe"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask
ParameterTypeRequiredDescription
urlsstring[]YesArray with exactly 2 image URLs: [first_frame, last_frame]. Supports both HTTP URLs and asset references (e.g. "asset://asset-20260326-abc123")
videoInputModestringYesMust be "keyframe"
promptstringNoText description guiding the transition
durationstringNoVideo length: "4s" to "15s"; sd2.5 supports up to "30s" (default: "5s")
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.
seedintNoReproducibility seed: -1 or omit for random; 0–2147483647 for fixed.
generateAudiobooleanNoWhether to synthesize a synced audio track. Default true; pass false for silent video.

Multi-Reference

Use multiple reference images, videos, and audio files to guide generation. Uses videoInputMode: "reference". For sd2.5 this is the default reference sub-task; see Video Edit & Extend below to edit or continue an existing video instead.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "inputs": {
      "urls": [
        "https://example.com/ref1.jpg",
        "https://example.com/ref2.jpg"
      ],
      "videoUrls": ["asset://asset-motion-video"],
      "audioUrls": ["https://example.com/audio.mp3"],
      "prompt": "Character walks through a garden",
      "duration": "5s",
      "videoInputMode": "reference",
      "resolution": "1280x720"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask
ParameterTypeRequiredDescription
urlsstring[]NoReference image URLs (max 9). Supports both HTTP URLs and asset references (e.g. "asset://asset-20260326-abc123")
videoUrlsstring[]NoReference video asset:// URIs — upload via /api/v1/assets/upload first; external URLs are rejected. sd2: max 3 videos, each ≤15s. sd2.5: up to 10 videos, each 2–30s and ≤200MB, total reference duration ≤30s, 480p–4K inputs supported.
audioUrlsstring[]NoReference audio inputs. sd2 / sd2-fast / sd2-mini: up to 3 audio files, each 2–15s, total ≤15s — audio cannot be the only reference on these models (add at least one image or video). sd2.5: up to 10 files, ≤15MB and 2–30s each, total ≤30s, and audio-only input is supported.
videoInputModestringYesMust be "reference"
promptstringNoText description
durationstringNoVideo length: "4s" to "15s"; sd2.5 supports up to "30s" (default: "5s")
resolutionstringYesRequired for reference mode. Options: 720x720, 720x960, 960x720, 1280x720, 720x1280, 1280x540
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.
seedintNoReproducibility seed: -1 or omit for random; 0–2147483647 for fixed.
generateAudiobooleanNoWhether to synthesize a synced audio track. Default true; pass false for silent video.

Reference Constraints

  • Max 9 images, 3 videos, 3 audio files
  • Max 12 total files across all types
  • Each video/audio must be ≤ 15 seconds
  • Images must be at least 400px on the shortest side
  • sd2.5: max 30 images, 10 videos, 10 audio files, and 50 total; video and audio totals are each ≤ 30 seconds

Video Edit & Extend

sd2.5 splits reference-based generation into three sub-tasks. Omit mode for ordinary reference generation, set mode: "edit" to edit an existing video, or mode: "extend" to continue one. Both require at least one video in videoUrls and always output the source video's aspect ratio.

Put the video you want to operate on first in videoUrls and refer to it as "Video 1" in your prompt. edit forces the output length to match that first (source) video, which must be 4–30s, so any duration you send is ignored; billing uses the source-video duration as the output duration, plus all reference videos as input. extend accepts 1–3 clips stitched in order, and the duration you request is the output length of this generation (unrelated to the source length) — billed like ordinary reference generation.

The 2.0-family models (sd2 / sd2-fast / sd2-mini) also accept mode: "edit" and mode: "extend": they infer the operation from your prompt (describe the edit or continuation explicitly), the aspect ratio follows the source video, and duration stays under your control with ordinary billing.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2.5",
    "inputs": {
      "mode": "edit",
      "videoInputMode": "reference",
      "videoUrls": ["asset://asset-source-video"],
      "urls": ["https://example.com/annotation-at-1.2s.png"],
      "source_frame_timestamps_ms": [1200],
      "prompt": "Replace the marked object with a red umbrella",
      "outputResolution": "720p"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask
ParameterTypeRequiredDescription
modestringNo"edit" or "extend". Omit for ordinary reference generation.
videoUrlsstring[]YesAt least one video; the first is the source ("Video 1") — edit output length and billing follow it. Later videos are extra references (edit) or additional clips stitched in order (extend, max 3).
urlsstring[]NoOptional annotation/reference images.
source_frame_timestamps_msnumber[]Nomode "edit" only. One non-negative source-video timestamp in milliseconds per image in urls.
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.

Seedance 2.5 Draft to Final

Submit an sd2.5 task with inputs.draft: true to generate a native 480P draft. Do not send outputResolution or upscaleResolution: the API fixes draft output at 480P and rejects either field if supplied, including outputResolution: "480p". This works for text, keyframe, and reference generation, but not edit or extend. Each draft and final task is billed separately.

# 1. Create a 480P draft. Save both taskId and orderId from the response.
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"sd2.5","inputs":{"prompt":"A blue cube slowly rotates on a white background","duration":"4s","draft":true}}'

# 2. Poll by taskId until status is COMPLETED, then review the draft output.
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/jobs/queryTask?taskId=YOUR_DRAFT_TASK_ID"

# 3. Within 7 days of draft creation, generate the 1080P final from its SeeGen orderId.
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"sd2.5","inputs":{"draftSourceOrderId":"YOUR_DRAFT_ORDER_ID"}}'

Save the orderId returned by createTask (or retrieve it through queryTask), and wait until the draft task is COMPLETED. Submit a second sd2.5 task with inputs.draftSourceOrderId set to that orderId. The server verifies that the completed draft belongs to your account, reuses its original inputs and seed, and renders a native 1080P final. Other final inputs are ignored. A matching composition is expected, but pixel-identical output is not guaranteed.

Important rules

  • Which ID? Use the SeeGen orderId returned by createTask or queryTask, not the taskId. SeeGen looks up the draft's gateway task ID internally; do not send a draft_task.id field to this API.
  • Why are the fields different from ARK? In the provider API, resolution selects output quality and content[].draft_task.id identifies the draft. In SeeGen, resolution is an aspect-ratio preset and outputResolution selects quality for ordinary tasks. For a draft, omit outputResolution; for the final, send draftSourceOrderId. Do not copy a provider request into this API unchanged.
  • When? The draft must be COMPLETED and belong to your account. A draft can be used for final rendering for 7 days from its creation. After that, create a new draft.
  • What is reused? The final automatically uses the draft's prompt, reference media, duration, aspect ratio, seed, audio setting, and task type. Send only draftSourceOrderId in final inputs; other generation inputs are ignored. You may set a new top-level callBackUrl for the final task.
  • How is it billed? Draft and final are separate paid tasks: the draft uses the 480P rate, and the final uses the native 1080P rate. If the draft used input videos, final pricing uses those original input-video durations; the draft video itself is not counted as an additional input video.
ParameterTypeRequiredDescription
draftbooleanNosd2.5 only. Set true for a native 480P draft. Do not send outputResolution or upscaleResolution; the API chooses 480P. Edit, extend, and upscale are unavailable.
draftSourceOrderIdstringNoThe orderId of your own completed sd2.5 draft, returned by createTask or queryTask. Generates a native 1080P final from that draft.

Query draft and final relations

GET /api/v1/jobs/relations with an orderId. For a draft, the response lists all finals created from it. For a final, it returns the sourceDraft. Only orders belonging to the API key owner are visible.

# A draft orderId returns its finals, oldest first (including queued and failed tasks).
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/jobs/relations?orderId=YOUR_DRAFT_ORDER_ID&limit=20"

# A final orderId returns its source draft.
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/jobs/relations?orderId=YOUR_FINAL_ORDER_ID"

The response contains kind (draft or final), sourceDraft, finals, totalCount, and nextCursor. Each related task includes orderId, taskId, status, createTime, and a URL when completed. For drafts, pass nextCursor as cursor to load more finals (limit: 1–50). Deleted or archived orders are hidden.

{
  "orderId": "YOUR_DRAFT_ORDER_ID",
  "kind": "draft",
  "sourceDraft": null,
  "finals": [
    {
      "orderId": "FINAL_ORDER_ID",
      "taskId": "FINAL_TASK_ID",
      "status": "COMPLETED",
      "createTime": 1790192800000,
      "url": "https://example.com/final.mp4"
    }
  ],
  "totalCount": 1,
  "nextCursor": null
}

Seedance Parameters Reference

Complete reference of all inputs parameters for the sd2 / sd2-fast / sd2-mini / sd2.5 models.

ParameterTypeRequiredDescription
promptstringNoText description (required for text-to-video, optional for other modes). Max 20000 chars.
urlsstring[]NoImage URLs. Supports both HTTP URLs and asset references (e.g. "asset://asset-20260326-abc123"). Maps to uploadedUrls internally.
videoUrlsstring[]NoVideo reference asset:// URIs (reference mode only). Must be uploaded via /api/v1/assets/upload first — external URLs are rejected.
audioUrlsstring[]NoAudio reference URLs (reference mode only). Audio-only input is supported on sd2.5; sd2 / sd2-fast / sd2-mini require at least one image or video alongside audio.
durationstringNo"4s" to "15s" normally; sd2.5 supports up to "30s". Default: "5s"
resolutionstringNoAspect ratio: auto (default) | 720x720 | 720x960 | 960x720 | 1280x720 | 720x1280 | 1280x540
outputResolutionstringNoOutput tier: "480p", "720p" (default), "1080p", "2k", or "4k". Seedance 2.0 Pro and Seedance 2.5 support native 480p/720p/1080p/4k; Fast and Mini support native 480p/720p. Other tiers use automatic upscale. For a draft, omit this field; output is fixed at 480p.
videoInputModestringNo"keyframe" (default) or "reference"
modestringNoReference-mode sub-task (all Seedance models): omit for ordinary reference generation, "edit" to edit the first video in videoUrls, "extend" to continue 1-3 clips. See Video Edit & Extend.
source_frame_timestamps_msnumber[]Nosd2.5 Video Edit only: one non-negative millisecond timestamp per annotation image.
seedintNoRandom seed for reproducibility. -1 or omit for server-random. Same seed + same inputs yields a closely-matched output (not bit-identical due to GPU non-determinism). Range: -1 to 2147483647.
draftbooleanNosd2.5 only. Set true to generate a 480P draft; see Draft to Final.
draftSourceOrderIdstringNosd2.5 only. The orderId of a completed draft owned by this API account; generates a native 1080P final.
generateAudiobooleanNoWhether to synthesize an audio track (speech, SFX, background music) synced to the video. Default true. Set false to produce a silent video — slightly faster, useful if you plan to dub separately.
bitrateModestringNoOutput bitrate tier at the same resolution: "standard" (default) or "high". "high" preserves more detail and reduces banding / blocking at ~3-5× the file size — it does not change resolution or price.
upscaleResolutionstringNo(Deprecated) Legacy split field, still accepted for backward compatibility. New integrations should use outputResolution, which now takes "2k" / "4k" directly. If both are sent, upscaleResolution takes precedence — except on models with native 1080p (Seedance2 Pro / Seedance 2.5), where upscaleResolution:"1080p" resolves to native 1080p (billed at the native rate).

Top-level request fields: model (required), inputs (required), callBackUrl (optional webhook URL).

Seedance Assets

Seedance assets are images, videos, and audio files that go through the ByteDance Volcano review process before they can be used in Seedance video generation tasks. Upload an asset, wait for it to become ACTIVE, then use its asset:// URL in your Seedance task.

Note: Real human image and video assets require official review, usually completed within seconds. Once approved, they can be used directly as references. Without review, generation may fail.

Platform Content Precheck

When material review is enabled for your account, uploads are also reviewed by SeeGen across all generation models. Upload, status and list responses include a separate contentReview object. It does not replace the original supplier status.

Only ALLOW or DISABLED permits use under the platform policy; the original model and supplier requirements still apply. PROCESSING means wait and poll asset status. PENDING means request a review. BLOCK or NEEDS_REPHRASE means replace the material. SERVICE_UNAVAILABLE means the review could not finish and can be retried. Generation rejected by this precheck does not charge credits.

To review an older upload or retry a temporary review failure, send its numeric assetId to POST /api/v1/assets/review. The endpoint returns assetId and contentReview; it does not retry or change supplier review.

{ "assetId": 123, "contentReview": { "status": "PROCESSING", "reviewId": "review-id" } }

Upload Asset

Two ways to upload: send a local file directly (multipart/form-data), or submit a publicly accessible HTTPS URL. Either way the asset is processed and reviewed automatically.

Method A — direct file upload (multipart/form-data)

Send a local file with no image host required. The media type is detected from the file's bytes (the filename extension is not trusted). Allowed: images (jpg/png/webp/gif/bmp/tiff/heic), videos (mp4/mov), audio (wav/mp3). Max 50MB per file (image ≤30MB, video ≤50MB, audio ≤15MB).

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -F "file=@/path/to/photo.jpg" \
  -F "name=my-photo" \
  https://seegen.ai/api/v1/assets/upload
ParameterTypeRequiredDescription
filefileYesThe local file (multipart form field). Media type is detected from content.
namestringNoAsset name (max 64 characters)

Method B — by URL (application/json)

If the file is already hosted at a public HTTPS URL.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/photo.jpg",
    "type": "IMAGE",
    "name": "my-photo"
  }' \
  https://seegen.ai/api/v1/assets/upload
ParameterTypeRequiredDescription
urlstringYesPublicly accessible HTTPS URL of the file to upload
typestringYes"IMAGE", "AUDIO", or "VIDEO"
namestringNoAsset name (max 64 characters)

Upload Response

{
  "assetId": 123,
  "volcAssetId": "asset-20260326-abc123",
  "type": "IMAGE",
  "status": "PROCESSING",
  "failReason": null,
  "url": "https://example.com/photo.jpg",
  "name": "my-photo",
  "createdAt": 1711234567890
}

Query Asset Status

Poll an asset's review status. When status is PROCESSING, the endpoint automatically checks for updates from the review system.

# Query by asset ID (numeric)
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/assets/status?assetId=123"

# Query by volcAssetId (string)
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/assets/status?assetId=asset-20260326-abc123"

Asset Status Values

  • PROCESSING — under review, not yet usable
  • ACTIVE — supplier review passed; also check platform contentReview before use
  • FAILED — review failed, check failReason

List Assets

List your uploaded assets with optional filtering by type and status. Supports cursor-based pagination.

# List all assets
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/assets/list"

# Filter by type and status
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/assets/list?type=IMAGE&status=ACTIVE&limit=10"

# Paginate with cursor
curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/assets/list?cursor=100&limit=20"
ParameterTypeRequiredDescription
typestringNoFilter by type: "IMAGE", "AUDIO", or "VIDEO"
statusstringNoFilter by status: "NONE", "PROCESSING", "ACTIVE", or "FAILED"
cursornumberNoCursor for pagination (use nextCursor from previous response)
limitnumberNoItems per page, 1-50 (default: 20)

List Response

{
  "items": [
    {
      "assetId": 123,
      "volcAssetId": "asset-20260326-abc123",
      "type": "IMAGE",
      "status": "ACTIVE",
      "failReason": null,
      "url": "https://example.com/photo.jpg",
      "name": "my-photo",
      "width": 1920,
      "height": 1080,
      "size": 245000,
      "duration": null,
      "createdAt": 1711234567890
    }
  ],
  "nextCursor": 122
}

Using Assets in Tasks

Once an asset is ACTIVE, use its volcAssetId with the asset:// protocol in your task's URLs:

{
  "model": "sd2",
  "inputs": {
    "urls": ["asset://asset-20260326-abc123"],
    "prompt": "The person slowly looks up and smiles",
    "duration": "5s"
  }
}

MiniMax H3 / H3 Max / H3 Max Turbo

Use minimax-h3, minimax-h3-max, or minimax-h3-max-turbo with POST /api/v1/jobs/createTask. Poll /api/v1/jobs/queryTask with the returned taskId until COMPLETED or FAILED. Save taskId and orderId; do not create a second task if a query times out. callBackUrl is optional at the top level; SeeGen sends the terminal result after local processing finishes.

Upload local reference files with POST /api/v1/assets/upload?model=minimax-h3 (or the selected Max / Turbo ID), Authorization: Bearer <API_KEY>, and multipart form field file. Pass the returned url in inputs.urls, videoUrls, or audioUrls. References must belong to your account; external URLs and asset:// identifiers are not accepted.

A prompt is required for every mode. H3 supports 4–15 whole seconds at 768p or 2K; Max and Turbo support 5–15 seconds at 480p or 768p. Native audio is generated with the video. H3 and Max accept up to 9 reference images, 3 videos, and 3 audio clips within 12 files. Video and audio references each have a separate 15-second total limit and a 2–15-second per-clip limit. Audio references require at least one image or video. Turbo supports text, a first frame, and first/last frames only.

Images: JPEG, PNG, WEBP, up to 30 MB, 256–5760 pixels per side, aspect ratio 0.4–2.5. Videos: MP4/MOV, up to 50 MB, 256–5760 pixels per side, 23.976–60 fps. Audio: MP3/WAV, up to 15 MB. Upload validates the file bytes and records metadata before generation.

Credits = output seconds × output rate + sum of each input video’s duration rounded up to a whole second × input video rate + images beyond the free allowance × image rate. All input images count, including first/last frames. Audio references are free. Server-measured video duration determines the charged amount; client-supplied durations are ignored. Failed tasks refund the reserved credits.

Text to Video

MiniMax H3

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3",
  "inputs": {
    "videoInputMode": "keyframe",
    "prompt": "Dark cinematic VFX title sequence beginning with a single drop of glossy black liquid falling onto polished metal. Instead of splashing normally, the fluid violently spreads outward in branching organic tendrils. The camera tracks the liquid at macro scale as it races across the surface, splitting, reconnecting and forming increasingly complex structures. Thousands of black droplets suddenly leap upward against gravity and spiral into a living fluid vortex. The viscous material stretches into sharp ribbons and rapidly constructs the word \"VENON\", every letter constantly moving beneath a wet reflective surface as if the typography itself were alive. The camera performs a sinister close orbit while black tendrils whip between the letters and droplets float around the lens. The title suddenly loses structural stability and violently liquefies, exploding toward the camera as an enormous black fluid splash. Hyper-realistic viscosity, organic procedural motion, glossy reflections, dark premium theatrical VFX.",
    "duration": "15s",
    "outputResolution": "768p",
    "ratio": "16:9",
    "videoWorkflowTab": "text-to-video"
  }
}'

MiniMax H3 Max

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max",
  "inputs": {
    "videoInputMode": "keyframe",
    "prompt": "High-speed chase shot through an abandoned post-apocalyptic city in daylight. The camera races backward ahead of a lone runner while a gigantic dust tornado destroys the avenue behind them. Cars roll through the air, windows explode and debris fills the sky. The tornado suddenly tears an entire building facade apart. Thousands of concrete, steel and glass fragments are sucked upward, rotating violently inside the vortex. From above, the debris unexpectedly aligns into enormous floating letters spelling “SURVIVE.” The runner passes underneath as the tornado instantly destroys the word again. Massive disaster-movie VFX, intense speed, photorealistic debris and atmospheric destruction.",
    "duration": "15s",
    "outputResolution": "768p",
    "ratio": "16:9",
    "videoWorkflowTab": "text-to-video"
  }
}'

MiniMax H3 Max Turbo

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max-turbo",
  "inputs": {
    "videoInputMode": "keyframe",
    "prompt": "Cinematic 15-second 16:9 historical travel film of Angkor Wat, Cambodia, at sunrise. Create 15 rapid one-second shots with clean cuts and consistent visual continuity.Ultra-realistic Khmer architecture: five lotus-shaped towers, sandstone galleries, bas-relief carvings, monumental gates, stone corridors, reflecting pools, tropical gardens, palms, jungle, and moats. Golden-orange sunrise with pink-purple skies, morning mist, soft sun rays, and mirror-like water reflections.Shot sequence: aerial temple reveal; entrance glide; symmetrical five-tower hero shot; carved sandstone close-up; pond reflection tilt-up; orbit around the central sanctuary; bas-relief tracking; side view of towers; corridor glide; low-angle central tower; tropical garden pan; sunlight through temple openings; high aerial complex view; texture push-in; final aerial pull-back over Angkor Wat and the sunrise horizon.Smooth drone, push, pan, tilt, slide, and orbit movements. Stable architecture, realistic water, natural mist, cinematic depth of field, HDR documentary style, consistent warm color grading. No warping, flicker, text, or people.",
    "duration": "15s",
    "outputResolution": "768p",
    "ratio": "16:9",
    "videoWorkflowTab": "text-to-video"
  }
}'

Image to Video

MiniMax H3

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "Four adult men cross the tree-lined street together, then continue walking toward the park entrance at the end of the road. The camera follows them from behind as they move naturally, with gentle wind through the trees and parked cars along the street. Keep the same hand-drawn animated style, daytime lighting, and calm cinematic mood. No text or watermark.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "image-to-video",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp"
    ]
  }
}'

MiniMax H3 Max

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "15-second, 16:9 realistic 3D animated action-comedy. Use @Image1 as the only reference. Keep the same black cat rider, dinosaur mount, long bread loaf, saddle, reins, canyon desert, harsh afternoon sunlight, and character scale. No text, subtitles, watermark, extra characters, scene changes, fades, or black ending.The dinosaur runs continuously from left-rear to right-front along a flat desert track. The cat stays in the saddle area throughout. The cat holds the bread against its chest; after the jump, the bread flies in a clear arc toward the dinosaur’s mouth, and the dinosaur catches and eats it.0–3s: Low side tracking shot. The dinosaur runs fast, kicking up dust. The cat grips the bread and shouts, “Slow down! My bread is gonna fly off!”3–5s: Close-up of the dinosaur giving a secret mischievous smile. The cat notices and says, “Don’t you dare—”5–10s: The dinosaur deliberately jumps high on the flat path. The cat is thrown upward, and the bread slips free. In brief slow motion, the cat reaches for it as crumbs trail behind; the dinosaur catches it with a crisp “CHOMP!”10–13s: The dinosaur lands heavily. The cat falls back onto the saddle, crawls forward angrily, and yells, “You did that on purpose!” The dinosaur chews innocently.13–15s: Dynamic split-screen close-up: the cat is shocked while the dinosaur grins. A final crumb lands on the cat’s nose; the cat cross-eyes at it.Natural high-speed footfalls, wind, leather creaks, reins, dust impacts, bread crunching, and dialogue. No music.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "image-to-video",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp"
    ]
  }
}'

MiniMax H3 Max Turbo

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max-turbo",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "The pale crust turning golden brown.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "image-to-video",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp"
    ]
  }
}'

First and Last Frames

MiniMax H3

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "15-second, 16:9 hyper-realistic dark science-fantasy naval battle. Stormy ocean, black cliffs, a massive coastal fortress, realistic water physics, violent explosions, and blockbuster cinematic scale.0–2s: Extreme wide shot. Three armored battleships fire at the ocean below the fortress. Explosions send water high into the air as a huge shadow moves beneath the waves.2–4s: A single colossal humanoid water giant erupts beside the lead ship and smashes it in half.4–6s: The same water giant sweeps one arm through the remaining two ships, forcing them into a catastrophic collision and explosion. No other water figures appear.6–9s: The water giant collapses into the sea. An enormous ancient armored leviathan rises from the ocean, dwarfing the fortress.9–12s: The leviathan tears the fortress from the cliff with one armored forelimb and crushes it, sending debris into the sea.12–15s: A huge wave swallows the last debris. At 13 seconds, the water settles around the shattered empty coastline. At 14 seconds, sunlight breaks through the storm clouds and illuminates the victorious leviathan. Hold the final shot.Audio: thunderous artillery, waves, steel destruction, orchestral percussion and brass; at 13 seconds, percussion stops and one deep sustained chord remains.One water giant only, one leviathan only. Keep scale, ships, geography, and lighting consistent. No dialogue, subtitles, text, logos, cartoon style, or abrupt ending.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "first-last-frame",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.webp"
    ]
  }
}'

MiniMax H3 Max

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "A smooth cinematic time-lapse from daytime to night. The sun slowly moves across the sky as clouds drift naturally overhead. Ships pass through the water in the distance. As the sky darkens, streetlights and building windows gradually illuminate one by one, creating a warm evening city glow. Keep the camera static, with realistic lighting transitions and natural continuous motion.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "first-last-frame",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.webp"
    ]
  }
}'

MiniMax H3 Max Turbo

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max-turbo",
  "inputs": {
    "videoInputMode": "keyframe",
    "ratio": "adaptive",
    "prompt": "A cinematic seaside scene. The little girl holds a camera and takes photos of a young seagull near the shore. A larger mother seagull then swoops down, and the young seagull follows her into the sky. The girl runs along the rocky beach after them, holding her camera.Keep the same 3D anime character, lavender twin-tail hair, black gothic outfit, sunny ocean, and bright blue sky. Smooth camera tracking, natural bird flight, no text or watermark.",
    "duration": "15s",
    "outputResolution": "768p",
    "videoWorkflowTab": "first-last-frame",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.webp"
    ]
  }
}'

Multiple References

MiniMax H3

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3",
  "inputs": {
    "videoInputMode": "reference",
    "prompt": "Use @Image1 as the first frame. The camera slowly pushes forward through the Japanese city street toward the giant white cat. The cat remains still at first, with subtle wind moving debris and clouds overhead.Then transition to @Image2: the cat opens its mouth as a bright cyan-blue energy glow builds inside, lighting its face and the nearby buildings.Transition to @Image3: the cat fires a powerful blue energy beam toward the right side of the street. The beam creates swirling shockwave rings and electric arcs. The building hit by the beam cracks, shakes, and collapses realistically into dust and debris. Keep the cat, city layout, anime style, lighting, and camera direction consistent throughout. No text, subtitles, or watermark.",
    "duration": "15s",
    "outputResolution": "768p",
    "ratio": "16:9",
    "videoWorkflowTab": "multi-reference",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_3.webp"
    ]
  }
}'

MiniMax H3 Max

Open API Playground
curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "minimax-h3-max",
  "inputs": {
    "videoInputMode": "reference",
    "prompt": "Create a 15-second Japanese anime comedy in one ordinary living room. Modern polished 2D animation, clean line art, vivid cel shading, cinematic lighting. Keep the same two characters, costumes, faces, room, and normal household vacuum cleaner throughout@Image1. No text, subtitles, logos, UI, or extra characters.0–5s: Character A stands dramatically as an enormous anime aura erupts around them: glowing energy, particles, flowing light, and wind in their hair and clothes. Character B watches calmly, holding a vacuum [email protected]–10s: From a closer side angle, Character B silently turns on the vacuum and points it at A@Image3. Only A’s aura is visibly sucked into the nozzle as a continuous stream of light and particles. A becomes confused; B remains unimpressed. Do not pull A’s body, hair, or clothes.10–15s: Wide static shot. A stands with no aura, devastated. B turns off the vacuum. After a short pause, the vacuum begins vibrating and the same huge supernatural aura erupts around it.@Image4 Keep the vacuum completely normal—no transformation. Both characters slowly stare at the awakened vacuum cleaner. Hold the final absurd image.Maximum three shots, stable character consistency, minimal camera movement, no dialogue.",
    "duration": "15s",
    "outputResolution": "768p",
    "ratio": "16:9",
    "videoWorkflowTab": "multi-reference",
    "urls": [
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_3.webp",
      "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_4.webp"
    ]
  }
}'

Parameters

ParameterTypeRequiredDescription
promptstringYesRequired non-empty scene description in every mode.
videoWorkflowTabstringNotext-to-video, image-to-video, first-last-frame, or multi-reference. Turbo does not support multi-reference. When omitted, mode is inferred from the inputs.
durationstring | integerNoDefault 5s. H3: 4–15 whole seconds. Max and Turbo: 5–15 whole seconds.
outputResolutionstringNoDefault 768p. H3: 768p or 2k. Max and Turbo: 480p or 768p. These are direct model output tiers; omit upscaleResolution.
ratiostringNo16:9, 9:16, 1:1, 4:3, 3:4, 21:9, or adaptive. Text-to-video requires a fixed ratio. First/last-frame generation follows the image aspect ratio.
urlsstring[]NoExactly one first-frame URL for image-to-video; exactly two ordered first/last URLs for frame mode; up to 9 reference images for multi-reference.
videoUrls / audioUrlsstring[]NoMulti-reference only, up to 3 files of each type. Use caller-owned uploaded material URLs.
promptExpansionModestringNoMax / Turbo: balanced (default) or quality.

Kling 4.0

Try Kling 4.0 in the API Playground

Generate 3–30 seconds at 720p or 1080p. Five workflows share one model: text, image, first and last frame, multiple keyframes, and image/video references. Native audio is generated from the prompt.

Upload references first with multipart POST /api/v1/assets/upload?model=kling-4.0. Use the returned url in media[].url. References must belong to your account. Replace the example URLs below with uploaded material URLs.

Use JPG/PNG images up to 30 MB, with both sides at least 300 px. Videos must be MP4/MOV, up to 300 MB, 3–30 seconds, 24–60 fps, and 700–4553 px per side with at most 8294400 pixels. References share limits of 10 images, 5 videos, 15 files and 30 seconds of input video. Keep reference IDs unique and use @Image1 or @Video1 in the prompt.

Without input video: 720p 54 and 1080p 80 credits per output second. With input video: 720p 32 and 1080p 48 credits per billable second. The server measures each input video and rounds its duration up to a whole second before summing. Billable seconds = output + max(rounded input total, output). Final credits are rounded up. The 30-second input limit uses actual durations before rounding.

Multi-keyframe uses first_frame, middle_frame and last_frame. Array order controls the order of middle frames. No explicit keyframe timestamp field is exposed; describe timing in the prompt.

Text to Video

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-4.0",
  "inputs": {
    "videoWorkflowTab": "text-to-video",
    "prompt": "A cinematic scene. Use the supplied references to guide the action and visual style.",
    "duration": "10s",
    "outputResolution": "1080p",
    "ratio": "16:9",
    "media": []
  }
}'

Multi-Keyframe

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-4.0",
  "inputs": {
    "videoWorkflowTab": "multi-keyframe",
    "prompt": "A cinematic scene. Use the supplied references to guide the action and visual style.",
    "duration": "10s",
    "outputResolution": "1080p",
    "media": [
      {
        "id": "Image1",
        "type": "first_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.png"
      },
      {
        "id": "Image2",
        "type": "middle_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.png"
      },
      {
        "id": "Image3",
        "type": "last_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_3.png"
      }
    ]
  }
}'

Image to Video

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-4.0",
  "inputs": {
    "videoWorkflowTab": "image-to-video",
    "prompt": "A cinematic scene. Use the supplied references to guide the action and visual style.",
    "duration": "10s",
    "outputResolution": "1080p",
    "media": [
      {
        "id": "Image1",
        "type": "first_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.png"
      }
    ]
  }
}'

First & Last Frame

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-4.0",
  "inputs": {
    "videoWorkflowTab": "first-last-frame",
    "prompt": "A cinematic scene. Use the supplied references to guide the action and visual style.",
    "duration": "10s",
    "outputResolution": "1080p",
    "media": [
      {
        "id": "Image1",
        "type": "first_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.png"
      },
      {
        "id": "Image2",
        "type": "last_frame",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_2.png"
      }
    ]
  }
}'

Multi-Reference

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "kling-4.0",
  "inputs": {
    "videoWorkflowTab": "multi-reference",
    "prompt": "A cinematic scene. Use the supplied references to guide the action and visual style.",
    "duration": "10s",
    "outputResolution": "1080p",
    "ratio": "adaptive",
    "media": [
      {
        "id": "Image1",
        "type": "refer_image",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_IMAGE_1.png"
      },
      {
        "id": "Video1",
        "type": "feature_video",
        "url": "https://static.seegen.ai/materials/YOUR_UPLOADED_VIDEO.mp4"
      }
    ]
  }
}'

Parameters

Use videoWorkflowTab to select the workflow, prompt for the instruction, duration as an integer or seconds string (default 10s), outputResolution as 720p or 1080p (default 1080p), and media as objects with id, type and url. ratio accepts 16:9, 9:16, 1:1 and 21:9 for text-to-video; multi-reference also accepts adaptive. Omit ratio for image/keyframe workflows. Voice and element references are not enabled.

Grok Imagine Video 1.5 / Lite

Use grok-imagine-video-1.5 or grok-imagine-video-1.5-lite with createTask and queryTask. Both support text-to-video and single-image-to-video with synced audio. First/last-frame and multi-reference generation are available only with 1.5.

Grok Imagine Video 1.5: Output costs 30 / 50 / 90 credits per second at 480p / 720p / 1080p, plus 5 credits per input image. Audio does not change the price. A 5-second 480p video with one input image costs 155 credits.

Grok Imagine Video 1.5 Lite: Output costs 8 / 12 / 50 credits per second at 480p / 720p / 1080p, plus 5 credits per input image. Audio does not change the price. A 5-second 480p video with one input image costs 45 credits.

Set SEEGEN_API_KEY to your API key. Replace example.com image URLs with your own publicly accessible HTTP(S) images. After creation, save taskId and use queryTask or Webhook Callback to receive the final result. Creation returns an asynchronous task, not a completed result.

Text to Video

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video-1.5",
  "inputs": {
    "prompt": "A blue toy cube slowly rotates on a white table",
    "duration": "5s",
    "outputResolution": "480p",
    "ratio": "16:9",
    "generateAudio": true
  }
}'

Image to Video

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video-1.5",
  "inputs": {
    "prompt": "A blue toy cube slowly rotates on a white table",
    "duration": "5s",
    "outputResolution": "480p",
    "generateAudio": true,
    "urls": [
      "https://example.com/first.png"
    ]
  }
}'

First & Last Frame1.5 only

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video-1.5",
  "inputs": {
    "prompt": "A blue toy cube slowly rotates on a white table",
    "duration": "5s",
    "outputResolution": "480p",
    "ratio": "16:9",
    "generateAudio": true,
    "urls": [
      "https://example.com/first.png",
      "https://example.com/last.png"
    ]
  }
}'

Multi-Reference1.5 only

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "grok-imagine-video-1.5",
  "inputs": {
    "prompt": "A blue toy cube slowly rotates on a white table",
    "duration": "5s",
    "outputResolution": "480p",
    "ratio": "16:9",
    "generateAudio": true,
    "urls": [
      "https://example.com/reference.png"
    ],
    "videoInputMode": "reference"
  }
}'

Parameters

ParameterTypeRequiredDescription
promptstringConditionalprompt: required for text-to-video; optional when supplying images.
durationstring | integerNoduration: an integer from 1 to 15 seconds, for example "5s" (default). outputResolution: 480p (default), 720p, or 1080p.
outputResolutionstringNo480p (default), 720p, or 1080p. 1.5 supports native 1080p for text or single-image input; first/last-frame and multi-reference support 480p and 720p. Lite 1080p is upscaled from 720p.
ratiostringNoratio: 16:9 (default), 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9, or 5:2. Single-image-to-video follows the input image ratio.
urlsstring[]NoPublic HTTP(S) image URLs. Omit for text-to-video; one image selects single-image-to-video. Lite accepts at most one image. With 1.5, two images select first/last-frame; set videoInputMode: "reference" for 1–14 reference images.
videoInputMode"keyframe" | "reference"No1.5 only: keyframe (default) or reference. Omit for Lite.
generateAudiobooleanNogenerateAudio: boolean, defaults to true. Set to false for silent output.
Try Grok Imagine Video 1.5Try Grok Imagine Video 1.5 Lite

Wan 3.0 Video & Wan 3.0 Video Prime

Wan 3.0 supports text-to-video, first-frame image-to-video, first-and-last-frame interpolation, mixed image/video/audio references, video editing, and video extension. wan3.0-video-prime accepts the same inputs as wan3.0-video and is optimized for faster generation.

Both aliases use SeeGen's asynchronous API: POST /api/v1/jobs/createTask returns a taskId; poll GET /api/v1/jobs/queryTask?taskId=... or provide callBackUrl for the terminal result.

Supported output: fixed durations of 2–30 seconds at native 480p, 720p, or 1080p, without a watermark. Automatic duration (duration: -1), 2K/4K output, and a custom watermark setting are not supported. Upload reference media first, then use the returned HTTPS URL.

Text-to-Video

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video-prime",
    "inputs": {
      "prompt": "A red paper boat glides across a calm pond at sunrise, locked camera, no text.",
      "duration": "5s",
      "outputResolution": "720p",
      "ratio": "16:9",
      "generateAudio": true,
      "promptExtend": true,
      "seed": 12345
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Video (First Frame)

Upload Wan media with multipart POST /api/v1/assets/upload?model=wan3.0-video (or the Prime alias), then use the returned owned HTTPS url in generation inputs. The returned assetId is only the SeeGen material record ID; do not send it in generation inputs.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -F "file=@/path/to/first-frame.webp" \
  "https://seegen.ai/api/v1/assets/upload?model=wan3.0-video"
{
  "assetId": 123,
  "type": "IMAGE",
  "status": "ACTIVE",
  "url": "https://static.seegen.ai/materials/api/.../first-frame.webp"
}
curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "inputs": {
      "urls": ["https://static.seegen.ai/materials/api/.../first-frame.webp"],
      "videoInputMode": "keyframe",
      "prompt": "The subject looks toward the camera as morning mist drifts past.",
      "duration": "5s",
      "outputResolution": "1080p",
      "generateAudio": true
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

First & Last Frame

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "inputs": {
      "urls": [
        "https://static.seegen.ai/materials/api/.../first-frame.webp",
        "https://static.seegen.ai/materials/api/.../last-frame.webp"
      ],
      "videoInputMode": "keyframe",
      "prompt": "A smooth continuous transition from sunrise to night.",
      "duration": "8s",
      "outputResolution": "720p"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Multi-Reference (Images, Videos & Audio)

Set videoInputMode: "reference", then send at least one supported media input through urls, videoUrls, or audioUrls; the prompt is optional. Refer to media by order as Image1, Image2, Video1, or Audio1. Upload each reference with the Wan multipart endpoint first, then use its returned owned HTTPS URL.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video-prime",
    "inputs": {
      "videoInputMode": "reference",
      "urls": ["https://static.seegen.ai/materials/api/.../reference-image.webp"],
      "videoUrls": ["https://static.seegen.ai/materials/api/.../source-video.mp4"],
      "audioUrls": ["https://static.seegen.ai/materials/api/.../reference-audio.mp3"],
      "prompt": "Use Image1 for identity, Video1 for motion, and Audio1 for timing.",
      "duration": "10s",
      "outputResolution": "720p",
      "ratio": "16:9",
      "generateAudio": true,
      "promptExtend": true
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Video Edit

Upload the source video first, then use mode: "edit" with videoInputMode: "reference". In the required prompt, describe how to edit Video1, the first video in videoUrls. The aspect ratio is Auto; choose the output duration. The reference-media limits and pricing below apply to both editing and extension.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "inputs": {
      "mode": "edit",
      "videoInputMode": "reference",
      "videoUrls": ["https://static.seegen.ai/materials/api/.../source-video.mp4"],
      "prompt": "Transform Video1 into clay animation, keeping the characters and camera movement.",
      "duration": "5s",
      "outputResolution": "720p",
      "ratio": "adaptive"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Video Extend

Use mode: "extend" with a source video and an instruction prompt, such as "Extend Video1 forward", followed by what happens next. Set ratio: "adaptive". The selected duration is the generated output length, not the source length plus the extension. File-to-video and webpage-to-video are not supported.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video-prime",
    "inputs": {
      "mode": "extend",
      "videoInputMode": "reference",
      "videoUrls": ["https://static.seegen.ai/materials/api/.../source-video.mp4"],
      "prompt": "Extend Video1 forward, continuing the motion of the character and keeping the scene consistent.",
      "duration": "5s",
      "outputResolution": "720p",
      "ratio": "adaptive"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Wan 3.0 Parameters Reference

Both Wan 3.0 models accept the same inputs fields. Choose an output duration from 2 to 30 seconds in whole-second increments.

ParameterTypeRequiredDescription
promptstringNoVideo description or instruction. Max 20,000 characters. Required for text-to-video, video editing, and video extension; optional for image-to-video, first/last-frame, and multi-reference. Editing and extension also require a video. Use Image1 / Video1 / Audio1 markers for ordered references.
modestringNoOmit or use "normal" for standard generation. Use "edit" or "extend" with a video and an instruction prompt. Both use reference mode, Auto aspect ratio, and a fixed output duration of 2–30 seconds.
urlsstring[]NoOwned HTTPS image URLs returned by the Wan upload endpoint. Omit for text-to-video; pass exactly 1 first frame, exactly 2 first/last frames, or up to 10 reference images in reference mode. Direct unowned external URLs and Seedance asset-protocol references are rejected.
videoUrlsstring[]NoUp to 5 owned HTTPS reference-video URLs returned by the Wan upload endpoint. Each clip must be 1–15s, all reference videos together must be at most 15s, and reference-video duration plus output duration must be at most 30s. Direct unowned external URLs and Seedance asset-protocol references are rejected.
audioUrlsstring[]NoUp to 5 owned HTTPS reference-audio URLs returned by the Wan upload endpoint. WAV or MP3; each clip must be 1–15s and all reference audio together must be at most 15s. Direct unowned external URLs and Seedance asset-protocol references are rejected.
videoInputModestringNoOmit for text-to-video; use "keyframe" for one first frame or two first/last frames; use "reference" for mixed image/video/audio references.
durationstringNoInteger-second string from "2s" to "30s". Default: "5s".
outputResolutionstringNoNative "480p", "720p" (default), or "1080p". 2K/4K output is not supported.
ratiostringNo"auto" or "adaptive" (both select adaptive framing), "16:9", "9:16", "1:1", "4:3", or "3:4". Used for text-to-video and multi-reference. Editing and extension always use Auto; image-to-video and first/last-frame output follows the keyframe images.
generateAudiobooleanNoGenerate synchronized speech, sound effects, and music. Default true. Set false for silent output; pricing is unchanged.
promptExtendbooleanNoLet the model enrich the prompt before generation. Default true.
seedintNo0–2147483647. Use the same seed and inputs for a closely matched retry.

Reference Media Requirements

  • Images: up to 10; JPEG/JPG/PNG (no transparency)/BMP/WebP; ≤20MB each; each side 240–8000px; aspect ratio up to 8:1
  • Videos: up to 5 MP4/MOV clips; ≤100MB each; 1–15s each, ≤15s total input, and input + requested output ≤30s; each side 240–4096px; aspect ratio up to 8:1
  • Audio: up to 5 uploaded WAV/MP3 HTTPS URLs returned by the Wan upload endpoint; ≤15MB each; 1–15s each and ≤15s total
  • First/last-frame mode cannot be mixed with reference image, video, or audio arrays

Choose a Model (Image)

Image models take a prompt (and optionally reference images) and return one image per task. Each request is billed per image using the selected model's tier (or a flat rate for Seedream Lite). Failed tasks are refunded automatically. There is no batch parameter — to generate multiple variations, call createTask once per image.

ModelT2II2I (Edit)Multi-RefMax Resolution2k / medium
seedream-v5.5✓✓up to 102ksee Pricing
seedream-v5.0-flash✓✓up to 102kflat rate
gpt-image-2.5-flare✓✓up to 104ksee Pricing
gpt-image-2.5-sunburst✓✓up to 104ksee Pricing
gpt-image-2✓✓up to 104ksee Pricing
nano-banana-2.1✓✓up to 144ksee Pricing
nano-banana-2✓✓up to 104ksee Pricing
nano-banana-pro✓✓up to 104ksee Pricing
seedream-v5.0-lite✓✓up to 104kflat rate
seedream-v5.0-pro✓✓up to 102ksee Pricing
qwen-image-2.1-pro✓✓up to 102kflat rate
qwen-image-2.1-turbo✓✓up to 102kflat rate

GPT Image 2.5 Flare / Sunburst

GPT Image 2.5 Flare focuses on fast image generation and editing. Sunburst focuses on detailed image generation and precise edits. Both work with text prompts and reference images.

Available in the workspace, Playground, and API. Both models accept the parameters below: medium/high/xhigh/max quality, 1K/2K/4K presets, and PNG/JPEG/WebP output. Actual dimensions may differ from the preset. JPEG and WebP are converted from the generated PNG without resizing. Each API task generates one image.

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-flare",
    "inputs": {
      "prompt": "A neon-lit cyberpunk alley at midnight, photoreal",
      "quality": "medium",
      "resolution": "2k",
      "aspectRatio": "16:9"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

Pass 1–10 reference images via urls. The model will use them as visual context for the edit described in prompt.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-sunburst",
    "inputs": {
      "urls": ["https://example.com/portrait.jpg"],
      "prompt": "Restyle as oil painting",
      "quality": "medium",
      "resolution": "1k",
      "aspectRatio": "1:1"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

GPT Image 2.5 Parameters Reference

Complete reference of all inputs parameters for the gpt-image-2.5-flare model.

ParameterTypeRequiredDescription
promptstringYesText description of the image to generate, or the edit to apply when urls is provided.
urlsstring[]NoReference image URLs for image-to-image (edit) mode (1–10 images). Omit for text-to-image. Public HTTPS URLs accepted.
qualitystringNo"medium" / "high" / "xhigh" / "max". Default: "medium". Credits scale by tier (see Pricing).
resolutionstringNo"1k" / "2k" / "4k". Default: "2k". Credits scale by tier (see Pricing).
aspectRatiostringNo"1:1" / "16:9" / "9:16" / "4:3" / "3:4". Default: "1:1".
outputFormatstringNo"png" / "jpeg" / "webp". Default: "png".

Input Image Requirements (image-to-image mode)

  • Up to 10 reference images per task
  • Max file size 50 MB per image
  • Shortest side ≥ 256px
  • Aspect ratio between 1:3 and 3:1
  • Formats: JPEG, JPG, PNG, WEBP

Credits per image scale by resolution and quality — see the Pricing section for the full table.

GPT Image 2

OpenAI's GPT Image 2 model for high-quality text-to-image and image-to-image editing. One workflow handles both modes — pass urls to switch into edit mode automatically. Output is delivered via R2 in the format you request (PNG / JPEG / WEBP). Model name: gpt-image-2.

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "inputs": {
      "prompt": "A neon-lit cyberpunk alley at midnight, photoreal",
      "quality": "medium",
      "resolution": "2k",
      "aspectRatio": "16:9"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

Pass 1–10 reference images via urls. The model will use them as visual context for the edit described in prompt.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "inputs": {
      "urls": ["https://example.com/portrait.jpg"],
      "prompt": "Restyle as oil painting",
      "quality": "medium",
      "resolution": "1k",
      "aspectRatio": "1:1"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

GPT Image 2 Parameters Reference

Complete reference of all inputs parameters for the gpt-image-2 model.

ParameterTypeRequiredDescription
promptstringYesText description of the image to generate, or the edit to apply when urls is provided.
urlsstring[]NoReference image URLs for image-to-image (edit) mode (1–10 images). Omit for text-to-image. Public HTTPS URLs accepted.
qualitystringNo"medium" / "high". Default: "medium". Credits scale by tier (see Pricing).
resolutionstringNo"1k" / "2k" / "4k". Default: "2k". Credits scale by tier (see Pricing).
aspectRatiostringNo"1:1" / "16:9" / "9:16" / "4:3" / "3:4". Default: "1:1".
outputFormatstringNo"png" / "jpeg" / "webp". Default: "png".

Input Image Requirements (image-to-image mode)

  • Up to 10 reference images per task
  • Max file size 50 MB per image
  • Shortest side ≥ 256px
  • Aspect ratio between 1:3 and 3:1
  • Formats: JPEG, JPG, PNG, WEBP

Credits per image scale by resolution and quality — see the Pricing section for the full table.

Nano Banana 2.1 / 2

Nano Banana 2.1 and 2 share text-to-image and image editing, 1K / 2K / 4K output, and 10 aspect ratios. Default output: 2K, square (1:1). Select a version to update the model ID, examples, and reference image limits below. Existing nano-banana-2 requests remain supported.

Version

Version 2.1: use model nano-banana-2.1 with up to 14 reference images. Version 2.1 supports 14 references; version 2 supports 10.

Each API task returns one image and is billed by resolution.

Try Nano Banana 2.1 in the API Playground

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2.1",
    "inputs": {
      "prompt": "A neon-lit cyberpunk alley at midnight, photoreal",
      "resolution": "2k",
      "aspectRatio": "16:9"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

Pass 1–14 reference images via urls. Describe the changes and the details to preserve in prompt.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2.1",
    "inputs": {
      "urls": ["https://example.com/portrait.jpg"],
      "prompt": "Restyle as oil painting",
      "resolution": "2k",
      "aspectRatio": "1:1"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Nano Banana 2.1 Parameters Reference

Complete reference of all inputs parameters for the nano-banana-2.1 model.

ParameterTypeRequiredDescription
promptstringYesText description of the image to generate, or the edit to apply when urls is provided.
urlsstring[]No1–14 public HTTP(S) reference image URLs for editing. Omit for text-to-image.
resolutionstringNo"1k" / "2k" / "4k". Default: "2k". Credits scale by tier (see Pricing).
aspectRatiostringNo"1:1" / "16:9" / "9:16" / "4:3" / "3:4" / "3:2" / "2:3" / "4:5" / "5:4" / "21:9". Default: "1:1".

Input Image Requirements (image-to-image mode)

  • Up to 14 reference images per task
  • Max file size 50 MB per image
  • Shortest side ≥ 256px
  • Aspect ratio between 1:3 and 3:1
  • Formats: JPEG, JPG, PNG, WEBP

Credits per image scale by resolution — see the Pricing section for the full table.

Nano Banana Pro

Nano Banana Pro is a high-fidelity image model with broader aspect-ratio coverage than gpt-image-2 — adds portrait/landscape presets (3:2, 2:3, 4:5, 5:4) and cinematic 21:9. One workflow handles both modes — pass urls to switch into edit mode automatically. Model name: nano-banana-pro.

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-pro",
    "inputs": {
      "prompt": "A neon-lit cyberpunk alley at midnight, photoreal",
      "resolution": "2k",
      "aspectRatio": "16:9"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

Pass 1–10 reference images via urls. The model will use them as visual context for the edit described in prompt.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-pro",
    "inputs": {
      "urls": ["https://example.com/portrait.jpg"],
      "prompt": "Restyle as oil painting",
      "resolution": "2k",
      "aspectRatio": "1:1"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Nano Banana Pro Parameters Reference

Complete reference of all inputs parameters for the nano-banana-pro model.

ParameterTypeRequiredDescription
promptstringYesText description of the image to generate, or the edit to apply when urls is provided.
urlsstring[]NoReference image URLs for image-to-image (edit) mode (1–10 images). Omit for text-to-image. Public HTTPS URLs accepted.
resolutionstringNo"1k" / "2k" / "4k". Default: "2k". Credits scale by tier (see Pricing).
aspectRatiostringNo"1:1" / "16:9" / "9:16" / "4:3" / "3:4" / "3:2" / "2:3" / "4:5" / "5:4" / "21:9". Default: "1:1".

Input Image Requirements (image-to-image mode)

  • Up to 10 reference images per task
  • Max file size 50 MB per image
  • Shortest side ≥ 256px
  • Aspect ratio between 1:3 and 3:1
  • Formats: JPEG, JPG, PNG, WEBP

Credits per image scale by resolution — see the Pricing section for the full table.

Seedream 5.5

Generate and edit images at 1K or 2K with 15 aspect ratios and up to 10 reference images. Model name: seedream-v5.5. Each task generates one image.

Text-to-Image

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-v5.5",
  "inputs": {
    "prompt": "Photorealistic close-up of a young adult woman with a short black bob, wispy bangs, dark brown eyes, and naturally rosy cheeks. She holds a bitten red-and-yellow apple against her slightly parted lips, looking directly into the camera with a curious, candid expression. Pink-and-white gingham blouse with delicate white lace ruffles and a small pale pink hair bow.\nHer face fills the center-right of the frame, with her hand and apple prominent in the lower-left foreground. Softly blurred pastel interior with colorful decorations and glowing round pendant lights. Gentle natural daylight, warm creamy tones, subtle film grain, visible skin texture, shallow depth of field, nostalgic Japanese lifestyle photography, intimate snapshot composition. no text or watermark.",
    "resolution": "2k",
    "aspectRatio": "16:9"
  }
}'

Image-to-Image (Editing)

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-v5.5",
  "inputs": {
    "prompt": "Replace the brown tabby cat with an orange tabby cat. Preserve its pose, expression, size, and position. Keep the fish, water splashes, lighting, background, and composition unchanged.",
    "resolution": "2k",
    "aspectRatio": "16:9",
    "urls": [
      "https://static.seegen.ai/web-static/public/api-playground/seedream-5-5/v1/seedream-5-5-original-image.webp"
    ]
  }
}'

Seedream 5.5 Parameters Reference

ParameterTypeRequiredDescription
promptstringYesImage description or edit instruction.
urlsstring[]No1–10 public HTTPS reference image URLs. Omit for text-to-image.
resolution"1k" | "2k"No"1k" or "2k". Default: "2k". 4K is not supported.
aspectRatiostringNo"1:1", "1:2", "2:1", "1:3", "3:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "9:21", or "21:9".

1K: 10 credits per image; 2K: 17 credits per image. Each reference image costs 4 credits, including the first.

Seedream 5.5

Seedream 5.0 Flash

Generate and edit images at 1K, 1.5K or 2K with 15 aspect ratios and up to 10 reference images. Model name: seedream-v5.0-flash. Each task generates one image.

Text-to-Image

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-v5.0-flash",
  "inputs": {
    "prompt": "A single plain cobalt blue ceramic mug centered on a light gray studio background, soft natural lighting, product photography, no text.",
    "resolution": "1.5k",
    "aspectRatio": "1:1",
    "outputFormat": "png"
  }
}'

Image-to-Image (Editing)

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "seedream-v5.0-flash",
  "inputs": {
    "prompt": "Change only the blue mug to bright red. Preserve the composition, shape, background, shadows and lighting.",
    "resolution": "2k",
    "aspectRatio": "1:1",
    "outputFormat": "png",
    "urls": [
      "https://static.seegen.ai/web-static/public/api-playground/seedream-v5-flash/v1/text-to-image.webp"
    ]
  }
}'

Seedream 5.0 Flash Parameters Reference

ParameterTypeRequiredDescription
promptstringYesImage description or edit instruction.
urlsstring[]No1–10 public HTTPS reference image URLs. Omit for text-to-image.
resolution"1k" | "1.5k" | "2k"No"1k", "1.5k" or "2k". Default: "2k". 4K is not supported.
aspectRatiostringNo"1:1", "1:2", "2:1", "1:3", "3:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "9:21", or "21:9".
outputFormat"png" | "jpeg"No"png" or "jpeg". Default: "png".

8 credits per generated image at every resolution. Reference images are free.

Seedream 5.0 Flash

Seedream 5.0 Lite

Fast text-to-image and image-to-image generation at 2K or 4K with 15 aspect ratios. Model name: seedream-v5.0-lite.

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-v5.0-lite",
    "inputs": {
      "prompt": "A cinematic product photograph, soft studio light",
      "resolution": "2k",
      "aspectRatio": "1:1"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-v5.0-lite",
    "inputs": {
      "prompt": "Restyle the references as a premium fashion campaign",
      "urls": ["https://example.com/reference.png"],
      "resolution": "4k",
      "aspectRatio": "3:4"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Seedream 5.0 Lite Parameters

ParameterTypeRequiredDescription
promptstringYesImage description or edit instruction.
urlsstring[]No1–10 public HTTPS reference image URLs. Omit for text-to-image.
resolutionstringNo"2k" or "4k". Default: "2k".
aspectRatiostringNo"1:1", "1:2", "2:1", "1:3", "3:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "9:21", or "21:9".

Seedream 5.0 Pro

High-fidelity generation and editing at 1K or 2K. Model name: seedream-v5.0-pro.

Text-to-Image

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-v5.0-pro",
    "inputs": {
      "prompt": "Editorial portrait with dramatic rim lighting",
      "resolution": "2k",
      "aspectRatio": "3:4"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Image-to-Image (Editing)

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-v5.0-pro",
    "inputs": {
      "prompt": "Turn the reference into a cinematic poster",
      "urls": ["https://example.com/reference.png"],
      "resolution": "2k",
      "aspectRatio": "1:2"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Seedream 5.0 Pro Parameters

ParameterTypeRequiredDescription
promptstringYesImage description or edit instruction.
urlsstring[]No1–10 public HTTPS reference image URLs. Omit for text-to-image.
resolutionstringNo"1k" or "2k". Default: "1k". 4k is not supported.
aspectRatiostringNo"1:1", "1:2", "2:1", "1:3", "3:1", "2:3", "3:2", "3:4", "4:3", "4:5", "5:4", "9:16", "16:9", "9:21", or "21:9".

Qwen Image 2.1 Pro / Turbo

Use qwen-image-2.1-pro or qwen-image-2.1-turbo. Both support text-to-image and image editing through the same endpoint, with one output image per task.

Pro: 13 credits per image. Turbo: 5 credits per image. 1K and 2K cost the same. Prices include any active image promotion.

Set SEEGEN_API_KEY to your API key. Replace example.com image URLs with your own publicly accessible HTTP(S) images. After creation, save taskId and use queryTask or Webhook Callback to receive the final result. Creation returns an asynchronous task, not a completed result.

Text to Image

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "qwen-image-2.1-pro",
  "inputs": {
    "prompt": "A blue glass cube on a white table",
    "resolution": "1k",
    "aspectRatio": "1:1"
  }
}'

Image to Image

curl -X POST https://seegen.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SEEGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "qwen-image-2.1-pro",
  "inputs": {
    "prompt": "Restyle the reference image as a watercolor painting",
    "resolution": "1k",
    "aspectRatio": "1:1",
    "urls": [
      "https://example.com/reference.png"
    ]
  }
}'

Parameters

ParameterTypeRequiredDescription
promptstringYesprompt: required. Describe the image or the changes to your reference images.
resolution"1k" | "2k"Noresolution: 1k (default) or 2k. aspectRatio: 1:1 (default), 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 4:5, 5:4, or 21:9. Resolution represents a pixel-area tier, so rectangular output sides vary.
aspectRatiostringNoresolution: 1k (default) or 2k. aspectRatio: 1:1 (default), 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 4:5, 5:4, or 21:9. Resolution represents a pixel-area tier, so rectangular output sides vary.
urlsstring[]Nourls: optional array of up to 10 public HTTP(S) image URLs. Each image must be at most 10 MB, with each side between 1 and 8000 pixels and an aspect ratio between 1:10 and 10:1. Supported formats: JPEG, PNG, BMP, TIFF, WebP and GIF. Browser file uploads support JPEG, PNG, BMP, WebP and GIF; use a public URL for TIFF, which may not display a browser preview.

Request a transparent background in the prompt when needed. PNG transparency is preserved. To create multiple outputs, submit separate tasks; Playground can submit a batch of 1–4 tasks.

Try Qwen Image 2.1Qwen Image 2.1 Turbo

Video Upscaler

Create Task

POST /api/v1/upscale/create

# source.url accepts ANY https video URL — your own CDN, OR a file you first
# uploaded to us via /api/v1/assets/upload (pass the r2Url it returns). No need
# to declare which: we detect it. External URLs are validated; our own are trusted.
curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "source": { "type": "url", "url": "https://your-cdn.com/video.mp4" },
    "targetResolution": "2k",
    "callBackUrl": "https://your-server.com/webhook"
  }' \
  https://seegen.ai/api/v1/upscale/create
ParameterTypeRequiredDescription
source.typestringYesUse "url" for any source. ("uploadId" is a legacy alias kept for backward compatibility.)
source.urlstringNoWith type="url". Any https video URL — your own CDN, or the r2Url returned by /api/v1/assets/upload. For external URLs, http and private/internal IPs are rejected (SSRF guard); URLs on our own asset host skip that check.
source.r2UrlstringNoLegacy — only with type="uploadId" (backward compat). New integrations should use type="url".
targetResolutionstringYes"720p", "1080p", "2k", or "4k". Must be higher than the source resolution.
callBackUrlstringNoWebhook URL invoked once on terminal state (completed or failed).

Create Response

{
  "taskId": "n770mo4sh6rpi690ff3gwymx",
  "orderId": "ord_2026...",
  "status": "validating"
}

Query Status

GET /api/v1/upscale/query?taskId=...

Status progresses validating → processing → completed / failed.

curl -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/upscale/query?taskId=n770mo4sh6rpi690ff3gwymx"

Completed Response

{
  "taskId": "n770mo4sh6rpi690ff3gwymx",
  "status": "completed",
  "targetResolution": "2k",
  "creditsConsumed": 900,
  "result": {
    "url": "https://static.seegen.ai/standalone-upscale/results/...mp4",
    "probedDurationSeconds": 30,
    "probedSourceResolution": "1280x720"
  },
  "error": null,
  "createdAt": "2026-04-29T01:23:45.000Z",
  "finishedAt": "2026-04-29T01:35:01.000Z"
}

Failed Response

{
  "taskId": "n770mo4sh6rpi690ff3gwymx",
  "status": "failed",
  "creditsConsumed": null,
  "result": null,
  "error": {
    "code": "SOURCE_RESOLUTION_TOO_HIGH",
    "message": "Source 3840x2160 is not below target 4k"
  }
}

Limits

  • Source: https URL or your previously uploaded R2 URL
  • Duration up to 600 s (clips under 5 s are billed as 5 s)
  • File size ≤ 200 MB
  • Format: MP4 / MOV / WebM
  • Source resolution must be lower than target

Pricing

  • 720P: 17 credits/sec (5s = 85, 30s = 510)
  • 1080P: 25 credits/sec (5s = 125, 30s = 750)
  • 2K: 38 credits/sec (5s = 190, 30s = 1140)
  • 4K: 50 credits/sec (5s = 250, 30s = 1500)
  • 5-second minimum; failures refund credits automatically

Error Codes

Common errors you can act on. Other failures return a self-explanatory message field — read that before assuming the code is one of these.

CodeMeaning
INVALID_URLURL is malformed or not https
URL_NOT_REACHABLECouldn't fetch the URL — check it's public and reachable
UNSUPPORTED_MEDIA_TYPEFile is not a video, or not in MP4 / MOV / WebM
FILE_TOO_LARGESource exceeds 200 MB
DURATION_EXCEEDS_LIMITSource longer than 600 s
SOURCE_RESOLUTION_TOO_HIGHSource already at or above target — pick a higher target
INSUFFICIENT_CREDITSNot enough credits — top up and retry
ACCOUNT_FROZENAccount may not spend credits — contact support

Webhook Callback

Instead of polling, you can provide a callBackUrl to receive results automatically when a task completes or fails.

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2",
    "callBackUrl": "https://your-server.com/webhook/callback",
    "inputs": {
      "prompt": "A cat playing piano",
      "duration": "5s"
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Callback Payload

When the task finishes, we send a POST request to your URL with the same format as the queryTask response:

{
  "taskId": "task_abc123",
  "orderId": "order_abc123",
  "model": "sd2",
  "status": "COMPLETED",
  "creditsUsed": 200,
  "output": [
    {
      "url": "https://static.seegen.ai/videos/result.mp4",
      "width": 1280,
      "height": 720
    }
  ],
  "error": null,
  "createTime": 1711234567890,
  "completeTime": 1711234612345
}

Retry Policy: After the initial attempt, a non-2xx response or network error triggers up to 3 retries after 1s, 5s, and 30s. Handle repeated callbacks idempotently by taskId. Use queryTask as a fallback if no callback arrives; delivery is not guaranteed after retries are exhausted.

Response Format

createTask Response

{ "taskId": "task_abc123", "orderId": "order_abc123" }

queryTask Response

{
  "taskId": "task_abc123",
  "orderId": "order_abc123",
  "model": "sd2",
  "status": "COMPLETED",
  "creditsUsed": 200,
  "output": [
    {
      "url": "https://static.seegen.ai/videos/result.mp4",
      "width": 1280,
      "height": 720
    }
  ],
  "error": null,
  "createTime": 1711234567890,
  "completeTime": 1711234612345
}

credits Response

{
  "credits": 5000,
  "availableCredits": 4800
}

Error Handling

Status CodeMeaningAction
400Invalid parametersCheck the error message and fix your request
401Invalid or missing API keyCheck your Authorization header format
402Insufficient creditsOpen Credits to purchase credits or enable Auto Recharge, then retry the request
403Access deniedYou can only query your own tasks
404Task not foundVerify the taskId is correct
500Internal server errorCheck the error and task history before retrying a create request; a lost response may still have created a task.

Pre-submit validation (400 with a code)

Seedance and Wan 3.0 requests are checked against SeeGen's product rules before any credits are charged. When a check fails, createTask returns { "message": "...", "code": "..." } with HTTP 400; Wan errors may also include a path. No task is created or charged.

ParameterTypeRequiredDescription
EMPTY_CONTENTcodeNoNo prompt and no reference image / video / audio.
AUDIO_ONLY_NOT_SUPPORTEDcodeNoAudio is the only reference on sd2 / sd2-fast / sd2-mini. Add an image or video, or use sd2.5 (audio-only supported).
UNSUPPORTED_MODELcodeNoThe model alias is not supported. For Wan 3.0, use exactly "wan3.0-video-prime" or "wan3.0-video".
UNSUPPORTED_RESOLUTIONcodeNoThe selected output tier is unavailable. Wan 3.0 launch supports native 480p / 720p / 1080p only; 2K/4K and upscaleResolution are rejected.
DURATION_OUT_OF_RANGEcodeNoduration is not an integer within the model range (Wan 3.0 "2s"–"30s", sd2 family "4s"–"15s", sd2.5 "4s"–"30s"). Wan smart duration (-1) is not exposed.
INVALID_MEDIA_COMBINATIONcodeNoThe media does not match the selected mode (for example, the wrong keyframe count, media in text-to-video, no media in multi-reference, or mixed keyframe and reference inputs).
EDIT_SOURCE_VIDEO_REQUIRED / EXTEND_SOURCE_VIDEO_REQUIREDcodeNomode "edit" / "extend" was requested without a video in videoUrls.
EDIT_SOURCE_DURATION_INVALIDcodeNosd2.5 Video Edit: a video in the request is shorter than 4s or longer than 30s (ARK applies 4–30s to every video of an edit task).
TOO_MANY_REFERENCEScodeNoMore reference images / videos / audio clips than the model accepts (Wan 3.0 10 / 5 / 5, sd2 family 9 / 3 / 3, sd2.5 30 / 10 / 10; extend ≤3 videos).
REFERENCE_VIDEO_DURATION_INVALIDcodeNoA reference video exceeds the per-clip cap, the combined reference-video input exceeds its model cap, or the input-plus-output combination is invalid. For Wan 3.0: each clip 1–15s, total input ≤15s, and input + requested output ≤30s (15+15 is valid; 15+16 is rejected).
REFERENCE_AUDIO_DURATION_INVALIDcodeNoA known Wan 3.0 reference-audio duration exceeds the 1–15s per-clip range or makes the combined reference-audio input exceed 15s.
REFERENCE_IMAGE_INVALIDcodeNoA Wan 3.0 image is missing, belongs to another user, has the wrong media type, or violates a known size, format, dimensions, aspect-ratio, or no-transparency requirement.
REFERENCE_VIDEO_INVALIDcodeNoAn uploaded Wan 3.0 video violates the size, MP4/MOV format, dimensions, or aspect-ratio requirement.
REFERENCE_AUDIO_INVALIDcodeNoAn uploaded Wan 3.0 audio file violates the size or WAV/MP3 format requirement.
ASSET_NOT_FOUND / EXTERNAL_URL / READ_TIMEOUT / READ_FAILEDcodeNoA videoUrls entry could not be resolved to one of your uploaded assets, or its duration could not be read.

Full Examples

Complete workflow: upload an asset, wait for review, create a task with the approved asset, and poll for the result.

const API_KEY = process.env.API_KEY;
const BASE = "https://seegen.ai/api/v1";
const headers = {
  "Authorization": `Bearer ${API_KEY}`,
  "Content-Type": "application/json",
};

// 1. Upload asset and wait for review
// (To upload a LOCAL file instead of a URL, POST multipart/form-data with a "file"
//  field — no Content-Type header, no "type" — see the Upload Asset section above.)
async function uploadAndWaitForAsset(url, type = "IMAGE") {
  const res = await fetch(`${BASE}/assets/upload`, {
    method: "POST",
    headers,
    body: JSON.stringify({ url, type }),
    signal: AbortSignal.timeout(120_000),
  });
  if (!res.ok) throw new Error(`Upload failed: ${(await res.json()).message}`);
  const asset = await res.json();
  console.log(`Asset uploaded: ${asset.assetId}, status: ${asset.status}`);

  // Poll until review completes
  const deadline = Date.now() + 15 * 60_000;
  while (true) {
    const statusRes = await fetch(
      `${BASE}/assets/status?assetId=${asset.assetId}`,
      { headers, signal: AbortSignal.timeout(30_000) }
    );
    if (!statusRes.ok) throw new Error(`Asset query failed (${statusRes.status}): ${await statusRes.text()}`);
    const status = await statusRes.json();
    if (status.status === "ACTIVE") {
      console.log(`Asset approved: asset://${status.volcAssetId}`);
      return status.volcAssetId;
    }
    if (status.status === "FAILED") {
      throw new Error(`Asset review failed: ${status.failReason}`);
    }
    if (Date.now() >= deadline) throw new Error(`Review pending: query assetId=${asset.assetId} later`);
    await new Promise((r) => setTimeout(r, 3000));
  }
}

// 2. Create a task
async function createTask(inputs, callBackUrl) {
  const res = await fetch(`${BASE}/jobs/createTask`, {
    method: "POST",
    headers,
    body: JSON.stringify({
      model: "sd2",
      inputs,
      ...(callBackUrl && { callBackUrl }),
    }),
    signal: AbortSignal.timeout(120_000),
  });
  if (!res.ok) throw new Error(`[${res.status}] ${(await res.json()).message}`);
  return res.json();
}

// 3. Poll until done
async function waitForResult(taskId, timeoutMs = 900000) {
  const start = Date.now();
  while (Date.now() - start < timeoutMs) {
    const res = await fetch(
      `${BASE}/jobs/queryTask?taskId=${taskId}`,
      { headers, signal: AbortSignal.timeout(30_000) }
    );
    if (!res.ok) throw new Error(`Query failed for ${taskId} (${res.status}): ${await res.text()}`);
    const result = await res.json();
    if (result.status === "COMPLETED") return result;
    if (result.status === "FAILED") throw new Error(result.error);
    await new Promise((r) => setTimeout(r, 5000));
  }
  throw new Error(`Still processing: query taskId=${taskId} later`);
}

// Full workflow: upload → review → generate → result
async function main() {
  // Upload image and wait for review
  const volcAssetId = await uploadAndWaitForAsset(
    "https://example.com/photo.jpg", "IMAGE"
  );

  // Create task with approved asset
  const { taskId } = await createTask({
    urls: [`asset://${volcAssetId}`],
    prompt: "The person slowly looks up and smiles",
    duration: "5s",
  });
  console.log(`Task: ${taskId}`);

  // Wait for video
  const result = await waitForResult(taskId);
  console.log(`Video: ${result.output[0].url}`);
}

main().catch(console.error);

Need help? Join our Discord or contact us