Seedance Video Generation
Generate AI videos from text prompts or images using the skills.sh script bundled with this skill.
Instructions
- Read the user's request and determine:
- The prompt (required) — a scene description for the video - The mode — text-to-video (default) or image-to-video (if an image path/URL is provided) - Any optional overrides: --model, --ratio, --resolution, --duration, --camera-fixed, --output, --json-only
- Locate and run the CLI script via Bash. The script is
skills.shin this skill's directory:bash "<skill_directory>/skills.sh" --prompt "..." [--image PATH] [options]
- The script auto-loads ARK_API_KEY from .env in the project root - Use a timeout of at least 600000ms for the Bash call (video generation takes minutes) - Always quote the prompt
- Report the result:
- On success, the script prints the downloaded file path to stdout. Show it to the user. - On failure, relay the error message and suggest fixes (check API key, model, parameters).
Setup
Requires ARK_API_KEY in a .env file or exported in the shell. Get one at https://console.byteplus.com/ark/region:ark+ap-southeast-1/apikey
Python 3.7+ required. The Python SDK (byteplus-python-sdk-v2) installs automatically on first run.
Defaults
| Option | Default |
|---|---|
| Model | seedance-1-5-pro-251215 (switches to seedance-2-0-260128 after 24 Feb 2026) |
| Ratio | 16:9 |
| Resolution | 720p |
| Duration | 5 seconds |
| Camera fixed | true |
| Output dir | Current directory |
Available Models
seedance-2-0-260128— Seedance 2.0 (default after 24 Feb 2026 18:00)seedance-1-5-pro-251215— best quality + native audioseedance-1-0-pro-250528— best quality, no audioseedance-1-0-pro-fast-251015— 3x faster, good for draftsseedance-1-0-lite-t2v-250428— budget text-to-video onlyseedance-1-0-lite-i2v-250428— budget image-to-video only
Prompting Tips
Structure prompts as: Subject + Movement + Scene + Camera + Style
Good: "A hawk slowly gliding downward over a misty canyon, wings spread, cinematic tracking shot" Bad: "A bird flying"
For I2V: describe the motion, not the scene (the model can see the image).
The 1.5 Pro model supports native audio — describe sounds, music, or dialogue in the prompt.
Example Commands
# Text-to-video, high quality
skills.sh --prompt "A lighthouse beam sweeping across a stormy sea at night, cinematic" --resolution 1080p --duration 8
# Image-to-video
skills.sh --prompt "The flower slowly blooms, petals unfurling" --image flower.jpg --duration 6
# Quick draft
skills.sh --prompt "A robot dancing in a disco" --model seedance-1-0-pro-fast-251015 --resolution 480p --duration 3