seedance-video

Seedance Video Generation

Use the ByteDance Seedance-1.5-pro model (doubao-seedance-1-5-pro-251215) to generate videos from text or images.

Prerequisites

Install the SDK:

pip install 'volcengine-python-sdk[ark]'

Features

Text-to-Video: Generate videos from text prompts
Image-to-Video: Generate videos from a first-frame image
Custom Parameters: Supports setting duration, aspect ratio, watermark, etc.
Task Management: Create tasks, query status, auto-download

Usage

1. Quick Video Generation (Text-to-Video)

cd ~/.openclaw/workspace/skills/seedance-video
python3 scripts/generate_video.py "一只可爱的猫咪在草地上玩耍" --wait -o cat.mp4

2. Custom Parameters

python3 scripts/generate_video.py "日落时分的海边" \
  --duration 10 \
  --ratio 16:9 \
  --wait \
  -o sunset.mp4

3. Image-to-Video

python3 scripts/generate_video.py "猫咪动起来" \
  --image-url https://example.com/cat.jpg \
  --duration 5 \
  --wait

4. Create Task Only (No Waiting)

python3 scripts/generate_video.py "星空下的城市" -o starry.mp4
# 返回任务ID

Then query the status and download later:

python3 scripts/generate_video.py --status <task_id> --wait -o starry.mp4

Parameter Description

Parameter	Default Value	Description
`prompt`	Required	Video description prompt
`-o, --output`	output.mp4	Output file path
`-m, --model`	doubao-seedance-1-5-pro-251215	Model ID
`-d, --duration`	5	Video duration (seconds)
`-r, --ratio`	16:9	Aspect ratio (16:9, 9:16, 1:1, 4:3, etc.)
`--watermark`	false	Add watermark
`--return-last-frame`	false	Return the last frame image
`--image-url`	None	First-frame image URL (image-to-video)
`--wait`	false	Wait for video generation to complete and download
`--status`	None	Query the status of a specified task ID

API Key Configuration

You need to set the ARK_API_KEY or SEEDANCE_API_KEY environment variable.

Configuration Method (Recommended)

Copy the configuration template:

cp .canghe-skills/.env.example .canghe-skills/.env

Edit the .canghe-skills/.env file and fill in your API Key:

ARK_API_KEY=your-actual-api-key-here

Or use environment variables

export ARK_API_KEY="your-api-key"
# 或
export SEEDANCE_API_KEY="your-api-key"

Loading Priority

System environment variable (process.env)
Current directory .canghe-skills/.env
User home directory ~/.canghe-skills/.env

Prompt Optimization Suggestions

According to the Seedance-1.5-pro Prompt Guide：

Specific description: Describe details such as the scene, subject, action, environment, etc.
Clear style: Specify the photography style (e.g., "cinematic", "documentary style")
Lighting description: Describe the lighting conditions (e.g., "golden hour", "soft natural light")
Camera language: Describe camera movement (e.g., "slow push-in", "gimbal shot")

Example Prompts

一只金毛犬在秋天的公园里奔跑，金色落叶飘落，下午的阳光透过树叶，电影感镜头，稳定器拍摄，4K画质

一个机器人在未来城市的霓虹灯街道上行走，赛博朋克风格，雨夜，倒影，广角镜头，电影色调

Python API Usage

from volcenginesdkarkruntime import Ark
from scripts.generate_video import create_video_task, wait_for_video

# 初始化客户端
client = Ark(
    base_url="https://ark.cn-beijing.volces.com/api/v3",
    api_key="your-api-key"
)

# 创建视频任务
result = create_video_task(
    client=client,
    model="doubao-seedance-1-5-pro-251215",
    prompt="一只鸟在天空中飞翔",
    duration=5
)

task_id = result["task_id"]

# 等待并获取结果
final_status = wait_for_video(client, task_id)
video_url = final_status["video_url"]

Notes

Asynchronous generation: Video generation is an asynchronous process and usually takes 30-60 seconds
Task retention time: Task data is retained for only 24 hours
Rate limiting: Pay attention to the account's RPM and concurrency limits
Video ratio: It is recommended that the video aspect ratio be close to the first-frame image ratio

Task Status

queued: Queued
running: Generating
succeeded: Success
failed: Failed