30s Videos, 50+ References. Try It Now
All models

Seedance 2.5

Video

Next-generation multimodal reference · native 1080p · up to 4K

What Is Seedance 2.5 API?

Overview

Seedance 2.5 (sd2.5) is ByteDance Seed's next-generation professional multimodal video generation model. It supports native 30-second video generation, up to 50 multimodal reference inputs, professional video editing, and multilingual video creation. Seedance 2.5 improves visual consistency, prompt understanding, camera movement, and cinematic video generation quality for professional content creation.

Available workflows:

  • Text to Video
  • Image to Video
  • First-Last Frame to Video
  • Multi-Reference to Video
  • Video Edit
  • Video Extend

Key Capabilities

  • 30-second video generation: Generate longer AI videos with improved consistency and storytelling.
  • 50 multimodal references: Support up to 30 images, 10 videos, and 10 audio files in a single generation.
  • Audio-only reference: Use audio as an independent input for music-driven and sound-aware video creation.
  • Professional video editing: Edit or extend an existing video with mode: "edit" / "extend" (Video Edit and Video Extend in the generator).
  • Enhanced realism and consistency: Improve character consistency, camera movement, and visual quality.
  • Multilingual generation: Support 10+ languages for global video creation.

Inputs

Seedance 2.5 supports image, video, and audio references for multimodal video generation.

ModalityLimitFormatsConstraints
ImagesUp to 30JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC, HEIFUp to 4K resolution, max 30 MB each
VideosUp to 10MP4, MOV480p–4K, max 200 MB each; each video 2–30s; total duration ≤ 30s
AudioUp to 10MP3, WAVMax 15 MB each; each file 2–30s; total duration ≤ 30s

Note: Total reference assets across all modalities must not exceed 50.

Parameters

ParameterTypeDefaultOptions / Description
promptstringText description of scenes, actions, camera movements, and creative instructions.
urlsarrayReference image URLs. Supports up to 30 images.
videoUrlsarrayReference video URLs, up to 10. Must be asset:// URIs uploaded through the asset API — external video URLs are rejected.
audioUrlsarrayUp to 10 audio URLs. Audio can be used as an independent reference input.
videoInputModeenum"keyframe" for image / first-last frame inputs, "reference" for multi-reference, edit, and extend.
modeenumOmit for standard generation. "edit" edits a reference video (pass exactly one and name it in the prompt); "extend" continues 1–3 reference videos in order.
durationstring"5s"Output duration, "4s"–"30s". Ignored when mode is "edit" — the output follows the source video length.
outputResolutionenum"720p"Output resolution tier: 480p, 720p, 1080p (all native) · 2k, 4k (automatic upscale). Higher tiers bill at a higher per-second rate.
resolutionenum"auto"Aspect-ratio preset: auto, 1280x720, 720x1280, 720x720, 960x720, 720x960, 1280x540. Required in reference mode, except mode "edit" / "extend", where the output follows the source video.
generateAudiobooleantrueGenerate a synced audio track (speech / SFX / BGM) alongside the video.
bitrateModeenum"standard"Bitrate tier at the same resolution: standard or high. Does not affect price.
seedinteger-1Optional seed for reproducible generation results. -1 generates a random seed.

Seedance 2.5 API Pricing

Seedance 2.5 API uses a pay-as-you-go credit system on SeeGen AI. The rates below use the $500 API Pack: 125,000 credits at $0.004 per credit.

ConditionRate10-sec clip
SD2.5 480p, no video input30 credits/sec (~$0.12/sec)~300 credits (~$1.20)
SD2.5 480p, with video input23 credits/sec (~$0.09/sec)Based on output + reference video duration
SD2.5 720p, no video input60 credits/sec (~$0.24/sec)~600 credits (~$2.40)
SD2.5 720p, with video input45 credits/sec (~$0.18/sec)Based on output + reference video duration
SD2.5 1080p, no video input150 credits/sec (~$0.60/sec)~1,500 credits (~$6.00)
SD2.5 1080p, with video input113 × (output + input) (~$0.45/sec)Based on output + reference video duration
SD2.5 2K, no video input90 credits/sec (~$0.36/sec)~900 credits (~$3.60)
SD2.5 2K, with video input45 × (output + input) + 30 × outputBased on output + reference video duration
SD2.5 4K, no video input100 credits/sec (~$0.40/sec)~1,000 credits (~$4.00)
SD2.5 4K, with video input45 × (output + input) + 40 × outputBased on output + reference video duration

Credit Calculation

1. No video input

credits = out × rate

2. With video input

credits = (out + in) × rate

out = output_seconds. in = the sum of ⌈duration⌉ for each input video. Each duration is rounded up, with a minimum input duration of ceil(out × 2 / 3). 480p, 720p, and 1080p are native output; 2K and 4K use upscaling-based pricing.

Run Seedance 2.5 API

Generate a video in two steps: create a generation task, then query the result using the returned taskId.

Create a Task

curl -X POST \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sd2.5",
    "inputs": {
      "videoInputMode": "reference",
      "urls": [
        "https://example.com/reference-image.jpg"
      ],
      "videoUrls": [
        "asset://asset-reference-video"
      ],
      "audioUrls": [
        "asset://asset-reference-audio"
      ],
      "prompt": "Create a cinematic scene using the provided character, motion, and audio references.",
      "duration": "10s",
      "resolution": "1280x720",
      "outputResolution": "720p",
      "generateAudio": true
    }
  }' \
  https://seegen.ai/api/v1/jobs/createTask

Query the Result

Use the returned taskId to check generation status and retrieve the completed video.

curl \
  -H "Authorization: Bearer $API_KEY" \
  "https://seegen.ai/api/v1/jobs/queryTask?taskId=YOUR_TASK_ID"
Reference videos must be uploaded as assets before generation and passed using an asset:// URI. See the asset upload guide.

Seedance 2.5 API Use Cases

Use Case30-Second GenerationMultimodal ReferencesVideo EditingMultilingual
Film & EntertainmentComplete scenes, long takes, and multi-shot storytellingCombine multiple actors, characters, props, and environmentsReplace actors or props, adjust expressions, and relight scenesCreate multilingual dubbed versions with lip-sync
Advertising & E-commerceGenerate TV commercials, brand stories, and product videosMaintain consistent products, brand identity, and visual styleReplace SKUs, models, labels, and localized contentProduce localized ads and product videos
Education & KnowledgeExplain a concept or recreate a historical scene in one videoVisualize scientific principles and maintain consistent educational assetsReplace presenters and refine demonstration stepsCreate multilingual lessons and narrated content
Industrial & EnterpriseDemonstrate complete workflows, operations, and maintenance proceduresKeep products, equipment, and production lines visually consistentUpdate labels, specifications, packaging, and product variantsGenerate multilingual training and instruction videos
Games & Virtual ContentCreate CG trailers, narrative scenes, and episodic contentCombine virtual characters, environments, and real-world referencesModify character age, expressions, styles, and themed appearancesProduce multilingual virtual hosts and interactive content

Seedance 2.5 can support professional video production across entertainment, advertising, education, industrial training, and virtual content workflows.

Seedance 2.5 vs Seedance 2.0

Seedance 2.5 expands Seedance 2.0 with longer video generation, more multimodal references, audio-only input, enhanced editing, and broader multilingual support.

FeatureSeedance 2.5Seedance 2.0
Maximum video durationUp to 30 secondsUp to 15 seconds
Reference inputsUp to 50 total: 30 images, 10 videos, and 10 audio filesUp to 12 total: 9 images, 3 videos, and 3 audio files
Audio-only referenceSupportedRequires an image or video reference
Multilingual generation10+ languages8+ languages and dialects
Video editingMore precise editing, extension, and local changesVideo editing and extension supported
Visual consistencyImproved consistency across longer and more complex scenesStrong consistency for shorter scenes

FAQs about Seedance 2.5 API

Is Seedance 2.5 API available on SeeGen AI?

Yes. Seedance 2.5 (sd2.5) is available through the SeeGen AI Playground and API using model ID sd2.5.

How long can Seedance 2.5 videos be?

Seedance 2.5 supports video generation up to 30 seconds in a single task.

What reference inputs does Seedance 2.5 support?

A single generation can use up to 50 reference assets: 30 images, 10 videos, and 10 audio files.

Can audio be used as the only reference?

Yes. Seedance 2.5 supports audio as an independent reference for music-driven, rhythm-matched, and sound-aware video generation.

How much does Seedance 2.5 API cost?

Seedance 2.5 uses pay-as-you-go credit pricing on SeeGen AI. Final cost depends on output resolution, output duration, and the duration of any reference videos. The rates shown above use the $500 API Pack rate of $0.004 per credit.

Can I use real human references with Seedance 2.5 API?

Yes. Real human image and video assets must first pass SeeGen AI’s official asset review before they can be used as references.