Seedance 2.5 Notes

Introduction

Seedance 2.5 (model: "seedance-2-5") shares the same endpoints as Seedance 2.0 (Base URL, authentication, asset library, task polling, and response format are all identical) — just switch model to seedance-2-5 when calling Create video generation. This page only lists what’s different in 2.5; everything not mentioned here (auth, asset library, polling, error codes, etc.) is identical to 2.0 — see:

Create video generation (base parameters, content array structure, mutual-exclusion rules)
Get video generation
Create asset group / Create asset
Get moderation result

Key differences

Item	Seedance 2.0	Seedance 2.5
`duration`	`-1` or `4`–`15` seconds	`-1` or `4`–`30` seconds
Reference images `reference_image`	up to 9	up to 30
Reference videos `reference_video`	up to 3, total ≤ 15s	up to 10
Reference audio `reference_audio`	up to 3	up to 10
Video editing / extension	Not supported	Supported, see “Rules” below
`output_format`	`mp4` only	Adds optional `mov`
`resolution` (currently available)	`480p` / `720p` / `1080p` / `4K`	`480p` / `720p` / `1080p`

The higher reference-asset limits and longer duration are upstream Seedance 2.5 model capabilities — see the Dreamina Seedance 2.5 tutorial. The higher duration and reference-asset limits are already available; resolution billing tiers follow the table above and the actual pricing page.

Rules

Seedance 2.5 has no explicit task-type parameter. Instead, it infers which of the tasks below a request belongs to from the role of media items in content and the intent expressed in the prompt; the allowed range for ratio and duration also differs by task type:

Task Type	Trigger Condition	`ratio`	`duration`
Text-to-video	No reference assets at all — text prompt only	A specific ratio or `adaptive`	`-1` or `4`–`30`
Reference-to-video	Provides `reference_image`, `reference_video`, or `reference_audio`, and the prompt carries no editing/extension intent	A specific ratio or `adaptive`	`-1` or `4`–`30`
Video editing	Provides a reference asset, and the prompt expresses an editing intent (add, remove, replace, modify, repaint, or restore)	Must be `adaptive`	Must be `-1`
Video extension	Provides a reference asset, and the prompt expresses an intent to extend forward/backward, continue, or carry on	Must be `adaptive`	`-1` or `4`–`30`
First-frame / first-last-frame video	Provides an image as `first_frame`, optionally paired with `last_frame`	Must be `adaptive`	`-1` or `4`–`30`

Request examples (standard text-to-video / image-to-video)

Identical to 2.0 usage — just swap model:

curl -X POST "https://api.gravitex.ai/v1/video/generations" \
  -H "Authorization: Bearer sk-xxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2-5",
    "content": [
      {"type": "text", "text": "Golden hour drone shot over rolling mountains, clouds rolling by"}
    ],
    "duration": 20,
    "resolution": "720p",
    "ratio": "16:9",
    "generate_audio": true
  }'

For parameter details, the content array structure, and asset:// mapping rules, see Create video generation. Next: Get video generation

API documentation

Chat & text

Safety & audio

Image Series

Video Series

Completions & Embeddings

Interface module

Introduction

Key differences

Rules

Request examples (standard text-to-video / image-to-video)

​Introduction

​Key differences

​Rules

​Request examples (standard text-to-video / image-to-video)

Introduction

Key differences

Rules

Request examples (standard text-to-video / image-to-video)