seedance-video — Claude Code skill: what it does, how to install
seedance-video generates video from text or images through AceDataCloud's Seedance API, and lets the agent poll the task, edit or extend a clip.
Apache-2.0 241 installs ★ 17 updated 2026-10-25
npx skills add AceDataCloud/Skills --skill seedance-video -a claude-codeWhat seedance-video does
The skill points the agent at AceDataCloud's endpoints. A generation request goes to POST /seedance/videos with a model, resolution, ratio and duration; the reply is a task id. Fetch the result by polling POST /seedance/tasks with that id. Resending the same request only to check progress bills you twice.
Model presets are part of the package. 2.5 covers clips up to 30 seconds and can edit or extend an existing one; it also accepts an audio-only reference. 2.0 keeps a referenced person or character consistent and renders up to 4k. Images get roles: first_frame, last_frame, reference_image. Edit and extend need reference_video plus ratio: "adaptive"; edit also wants duration: -1.
Calls fail without ACEDATACLOUD_API_TOKEN in the agent's environment and HTTPS reach to api.acedata.cloud, so the key belongs in the environment, not in a prompt. It has to be a service key; a platform-... management token will not work. Ace Data Cloud maintains this skill, and it is not an official listing from the Seedance publisher.
What it helps with
- turn a written scene into a short clip
- animate a still photo from a chosen first frame
- keep a character's face consistent across shots
- extend a finished clip or rework one segment
- build a transition between two stills
Example prompts for the agent
- Generate a 5-second clip of a street dancer, model
doubao-seedance-1-0-pro-250528,1080p,16:9 - Use
dancer.jpgas thefirst_frameand animate it: the person starts dancing - Extend this clip with
omni_reference_task_type: "extend"andratio: "adaptive"
How to install seedance-video
1Skills CLI (Vercel)
npx skills add AceDataCloud/Skills --skill seedance-video -a claude-code2Manually: put the files in place
~/.claude/skills/seedance-video/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/seedance-video/
1Skills CLI (Vercel)
npx skills add AceDataCloud/Skills --skill seedance-video -a codex2Manually: put the files in place
.agents/skills/seedance-video/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add AceDataCloud/Skills --skill seedance-video -a cursor2Manually: put the files in place
.cursor/skills/seedance-video/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install AceDataCloud/Skills skills/seedance-video2Manually: put the files in place
.github/skills/seedance-video/SKILL.md1Skills CLI (Vercel)
npx skills add AceDataCloud/Skills --skill seedance-video -a gemini-cli2Platform CLI
gemini skills install https://github.com/AceDataCloud/Skills --path skills/seedance-videoManually: put the files in place
.devin/skills/seedance-video/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add AceDataCloud/Skills --skill seedance-video -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about seedance-video
How do I install seedance-video in Claude Code?
Run npx skills add AceDataCloud/Skills --skill seedance-video -a claude-code. It writes the skill to ~/.claude/skills/seedance-video/SKILL.md.
What is seedance-video?
A skill that lets the agent generate video through AceDataCloud's Seedance API: text or image in, clip out, with edit and extend for footage you already have.
How is Seedance 2.5 different from 2.0?
2.5 goes up to 30 seconds and handles edit, extend and a pure-audio reference. 2.0 renders up to 4k and keeps a referenced person or character consistent.
Source text
SKILL.md · Apache-2.0
---
name: seedance-video
description: Seedance video by Ace Data Cloud. Generate text-to-video and image-to-video, use multimodal references, or edit and extend video. Use when the user asks for Seedance 2.0 or 2.5 video generation.
license: Apache-2.0
metadata:
author: acedatacloud
version: "1.0"
compatibility: Requires ACEDATACLOUD_API_TOKEN in the agent process environment and HTTPS access to api.acedata.cloud. See README.md for installation and key setup.
---
# Seedance by Ace Data Cloud
Generate AI dance and motion videos through AceDataCloud's Seedance (ByteDance) API.
## Install and connect from skills.sh
This integration is maintained by **Ace Data Cloud**; it is not an official Seedance publisher listing. Install only this Skill with `npx skills add acedatacloud/skills --skill seedance-video --yes`, or use the [public skills.sh page](https://skills.sh/acedatacloud/skills/seedance-video). Keep the slug `seedance-video` when selecting it.
Read the [illustrated first-run guide](README.md) ([简体中文](readme.zh-CN.md)) for registration, **Applications → General application**, API key access and price checks, and the environment variable location for your coding agent. For skills.sh installations, use this guide; any ClawHub-specific setup applies only to OpenClaw. The shared references linked here are online because focused installs do not include sibling `_shared/` folders.
## ClawHub / OpenClaw setup
1. Install the official Skill with `clawhub install @acedatacloud/seedance`. Check that the ClawHub publisher is **@acedatacloud**.
2. Sign in at [AceDataCloud → Applications](https://platform.acedata.cloud/console/applications), open **General application**, and check Seedance access, balance, and current pricing. Copy its API key, or select **Manage Keys → Create** for a separate key. Copy the key value without `Bearer `; a `platform-...` management token is not a service API key.
3. If you enable **Allowed APIs** restrictions, include `/seedance/videos` and `/seedance/tasks`. Advanced workflows below may call additional routes.
4. On the OpenClaw Gateway host, put `ACEDATACLOUD_API_TOKEN=your_api_key_here` in `~/.openclaw/.env`, then restart the Gateway and start a new session. For a direct terminal test, run `export ACEDATACLOUD_API_TOKEN="your_api_key_here"` in that shell. Never put the key in a prompt or commit an `.env` file.
5. Run the **Quick Start** below. If this workflow returns a task ID, query that same task until it finishes; do not submit the same billable request again just to check progress.
[Illustrated key setup, first run, and troubleshooting](https://github.com/AceDataCloud/Skills#first-run-with-clawhub-and-openclaw) · [Token details](https://github.com/AceDataCloud/Skills/blob/main/skills/_shared/authentication.md)
## Quick Start
```bash
curl -X POST https://api.acedata.cloud/seedance/videos \
-H "Authorization: Bearer $ACEDATACLOUD_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"model": "doubao-seedance-2-0-260128", "content": [{"type": "text", "text": "a dancer performing contemporary ballet in a misty forest"}], "async": true}'
```
> **Async:** See [async task polling](https://github.com/AceDataCloud/Skills/blob/main/skills/_shared/async-tasks.md). Poll via `POST /seedance/tasks` with `{"id": "..."}`.
This returns a task ID immediately. Poll for the result:
```bash
curl -X POST https://api.acedata.cloud/seedance/tasks \
-H "Authorization: Bearer $ACEDATACLOUD_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"id": "<task_id from above>"}'
```
## Models
### Seedance 2.5 (latest flagship)
| Model | Best For | Limits |
|-------|----------|--------|
| `doubao-seedance-2-5-260628` | Up to 30 seconds, pure-audio or multimodal reference, video edit/extend | `480p`, `720p`, `1080p`; 30 images / 10 videos / 10 audios / 50 total |
Use `omni_reference_task_type: "auto"`, `"edit"`, or `"extend"`. Edit and extend require `reference_video` and `ratio: "adaptive"`; edit also requires `duration: -1`. Optional `output_format` is `mp4` or `mov`.
### Seedance 2.0 (multimodal reference, up to 4k)
The 2.0 series adds multimodal reference inputs: real-person / character **image** references, reference **audio**, and reference **video** (see the workflows below).
| Model | Best For | Max resolution |
|-------|----------|----------------|
| `doubao-seedance-2-0-260128` | Highest quality, real-person/character reference, 4k output | `4k` |
| `doubao-seedance-2-0-fast-260128` | Faster 2.0 generation | `720p` |
| `doubao-seedance-2-0-mini-260615` | Lightweight / most cost-effective 2.0 | `720p` |
### Seedance 1.x
| Model | Type | Best For |
|-------|------|----------|
| `doubao-seedance-1-5-pro-251215` | Text+Image-to-Video | 1.5 flagship, audio support |
| `doubao-seedance-1-0-pro-250528` | Text+Image-to-Video | General-purpose, reliable quality |
| `doubao-seedance-1-0-pro-fast-251015` | Text+Image-to-Video | Faster Pro generation |
| `doubao-seedance-1-0-lite-t2v-250428` | Text-to-Video only | Lightweight text-to-video |
| `doubao-seedance-1-0-lite-i2v-250428` | Image-to-Video only | Lightweight image-to-video |
## Workflows
### 1. Text-to-Video
Pass a text content item in the `content` array.
```json
POST /seedance/videos
{
"model": "doubao-seedance-1-0-pro-250528",
"content": [
{"type": "text", "text": "a street dancer doing breakdancing moves in an urban setting"}
],
"resolution": "1080p",
"ratio": "16:9",
"duration": 5
}
```
### 2. Image-to-Video
Include an image content item (with an optional `role`) alongside the text.
```json
POST /seedance/videos
{
"model": "doubao-seedance-1-5-pro-251215",
"content": [
{"type": "text", "text": "the person starts dancing gracefully"},
{
"type": "image_url",
"role": "first_frame",
"image_url": {"url": "https://example.com/dancer.jpg"}
}
],
"resolution": "720p",
"duration": 5
}
```
Image roles:
- `first_frame` — image is used as the opening frame
- `last_frame` — image is used as the closing frame
- `reference_image` — image is used as a style / subject / real-person reference (Seedance 2.0 keeps the referenced person or character consistent)
Reference media (Seedance 2.x):
- `audio_url` — use `role: "reference_audio"`
- `video_url` — use `role: "reference_video"`
- 2.5 accepts pure-audio reference; 2.0 requires an image or video alongside audio
### 3. First-frame + Last-frame
Provide both a start and end frame image:
```json
POST /seedance/videos
{
"model": "doubao-seedance-2-0-260128",
"content": [
{"type": "text", "text": "smooth transition between two scenes"},
{"type": "image_url", "role": "first_frame", "image_url": {"url": "https://example.com/start.jpg"}},
{"type": "image_url", "role": "last_frame", "image_url": {"url": "https://example.com/end.jpg"}}
]
}
```
### 4. Real-person / character reference (Seedance 2.0)
Seedance 2.0 models (`doubao-seedance-2-0-260128`, `doubao-seedance-2-0-fast-260128`, `doubao-seedance-2-0-mini-260615`) can keep a **specific person or character** consistent across a brand-new scene. Pass one or more photos as `image_url` items with `role: "reference_image"` — the model preserves that subject's appearance. Up to 9 reference images are accepted.
```json
POST /seedance/videos
{
"model": "doubao-seedance-2-0-260128",
"content": [
{"type": "text", "text": "the same person walking through a neon-lit night market, cinematic"},
{"type": "image_url", "role": "reference_image", "image_url": {"url": "https://example.com/person.jpg"}}
],
"resolution": "1080p",
"duration": 8
}
```
### 5. Reference audio / video (Seedance 2.0)
2.0 models also accept reference **audio** (voice timbre, background music) and reference **video** (subject content, camera movement, motion, overall style). Add `audio_url` and/or `video_url` content items. Limits: up to 3 audio and 3 video references per request.
```json
POST /seedance/videos
{
"model": "doubao-seedance-2-0-260128",
"content": [
{"type": "text", "text": "a singer performing on stage, matching the reference voice and motion"},
{"type": "image_url", "role": "reference_image", "image_url": {"url": "https://example.com/person.jpg"}},
{"type": "audio_url", "role": "reference_audio", "audio_url": {"url": "https://example.com/voice.mp3"}},
{"type": "video_url", "role": "reference_video", "video_url": {"url": "https://example.com/motion.mp4"}}
],
"generate_audio": true
}
```
## Parameters
| Parameter | Values | Description |
|-----------|--------|-------------|
| `model` | see Models table | Model to use (required) |
| `content` | array | Input items: `text`, `image_url`, `audio_url` (2.0), `video_url` (2.0) (required) |
| `resolution` | `"480p"`, `"720p"`, `"1080p"`, `"4k"` | Output resolution. `4k` is `doubao-seedance-2-0-260128` (standard) only; `2-0-fast` / `2-0-mini` max out at `720p` (default: 720p for pro/2.0, 480p for lite) |
| `ratio` | `"16:9"`, `"4:3"`, `"1:1"`, `"3:4"`, `"9:16"`, `"21:9"`, `"adaptive"` | Aspect ratio (default: 16:9) |
| `duration` | `2` – `30`, or `-1` | 1.0: 2–12 (no auto); 1.5 Pro: 4–12/-1; 2.0: 4–15/-1; 2.5: 4–30/-1 |
| `frames` | 29–289 (must satisfy 25+4n) | Frame count — mutually exclusive with `duration` |
| `seed` | -1 to 4294967295 | Seed for reproducible results (-1 = random) |
| `generate_audio` | `true` / `false` | Generate audio (supported by `doubao-seedance-1-5-pro-251215` and the `doubao-seedance-2-0` series; other models ignore it) |
| `camerafixed` | `true` / `false` | Fix the camera position during generation |
| `watermark` | `true` / `false` | Add a watermark to the generated video |
| `return_last_frame` | `true` / `false` | Return the last frame of the generated video |
| `omni_reference_task_type` | `auto`, `edit`, `extend` | Seedance 2.5 task type |
| `output_format` | `mp4`, `mov` | Seedance 2.5 output format |
| `tools` | object[] | Seedance 2.5 web search tool, e.g. `[{"type":"web_search"}]` |
| `priority` | 0–9 | Seedance 2.5 queue priority |
| `safety_identifier` | string | Stable anonymous end-user ID (max 64 chars; do not send PII) |
| `execution_expires_after` | number | Task timeout threshold in seconds |
## Inline Parameter Syntax
You can also embed generation parameters directly in the text prompt using the `--param value` syntax:
```
A kitten yawning at the camera. --rs 720p --rt 16:9 --dur 5 --fps 24 --seed 42
```
Supported inline params: `--rs` (resolution), `--rt` (ratio), `--dur` (duration), `--frames`, `--fps` (24 only), `--seed`, `--cf` (camera_fixed), `--wm` (watermark).
## Gotchas
- Model names use the `doubao-*` convention (e.g. `doubao-seedance-1-0-pro-250528`) — old short names like `seedance-1.0` are not valid
- The `content` array replaces the old `prompt` + `image_url` fields; always use `content`
- Image and text scenarios are mutually exclusive per content item — each item has either `text` or `image_url`, not both
- `first_frame` and `last_frame` may be combined in one request, but `reference_image` is mutually exclusive with `first_frame` / `last_frame` — do not mix a reference image with first/last frames
- `generate_audio: true` is supported by `doubao-seedance-1-5-pro-251215` and the `doubao-seedance-2-0` series; other models ignore this field
- Lite models are split: `*-lite-t2v-*` only accepts text, `*-lite-i2v-*` only accepts image-to-video
- `audio_url` and `video_url` reference items are supported by Seedance 2.0 and 2.5; 2.5 also supports pure-audio reference
- Resolution options are `480p`, `720p`, `1080p`, and `4k` (`4k` is `doubao-seedance-2-0-260128` only; `2-0-fast` / `2-0-mini` max out at `720p`) — there is no 360p or 540p
- Duration is model-specific: 1.0 supports 2–12; 1.5 Pro supports 4–12/-1; 2.0 supports 4–15/-1; 2.5 supports 4–30/-1
- Task states use `"succeeded"` (not "completed") — check for this value when polling
> **MCP:** `pip install mcp-seedance` | Hosted: `https://seedance.mcp.acedata.cloud/mcp` | See [all MCP servers](https://github.com/AceDataCloud/Skills/blob/main/skills/_shared/mcp-servers.md)
© acedatacloud · Apache-2.0 · AceDataCloud/Skills