This post covers the Seedance 2.0 API — ByteDance’s multimodal AI video generation model, now accessible through EvoLink. The focus is on practical integration: three access methods, all three generation modes with code examples, the async task workflow, pricing model, and optimization techniques.
Model Capabilities Overview
Seedance 2.0 introduces several capabilities that distinguish it from previous-generation video models:
- Multimodal @-reference system: Up to 9 images + 3 video clips + 3 audio tracks as simultaneous input references per request
- Video-to-video editing: Modify specific elements in existing video while preserving overall structure and timing
- Frame-accurate audio synchronization: Auto-generated dialogue, sound effects, and background music aligned to individual frames
- Multi-shot narrative generation: Structured sequences with camera transitions maintaining subject consistency across cuts
- Six model variants: Standard and fast versions of each generation mode
API Architecture
The API uses an async task pattern appropriate for long-running generation operations:
POST /v1/videos/generations → immediate task ID response
GET /v1/tasks/{task_id} → status polling (pending → processing → completed/failed)
Alternatively: callback_url parameter for webhook notification
All six models share identical request/response schemas. Switching between modes requires only changing the model parameter.
Integration Path 1: Web Playground
No-code visual interface at evolink.ai. Exposes all three generation modes with a parameter configurator and cost calculator. Recommended for initial model evaluation and prompt development.
Integration Path 2: ClawHub Agent Skill
For OpenClaw / Claude Code environments. The ClawHub skill abstracts parameter management, task submission, status polling, and result delivery behind a natural language interface. Install, set EVOLINK_API_KEY, and describe what you want to generate.
Integration Path 3: Direct API
RESTful API with JSON payloads. Full programmatic control for production applications.
Task Creation
curl --request POST \
--url https://api.evolink.ai/v1/videos/generations \
--header 'Authorization: Bearer YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"model": "seedance-2.0-text-to-video",
"prompt": "A macro lens focuses on a green glass frog on a leaf. The focus gradually shifts from its smooth skin to its completely transparent abdomen, where a bright red heart is beating powerfully and rhythmically.",
"duration": 8,
"quality": "720p",
"aspect_ratio": "16:9",
"generate_audio": true
}'
Response:
{
"id": "task_abc123",
"status": "processing",
"estimated_time": 90
}
Status Polling
curl --request GET \
--url https://api.evolink.ai/v1/tasks/task_abc123 \
--header 'Authorization: Bearer YOUR_API_KEY'
Generation Modes
Text-to-Video
Prompt-only generation. Standard model: seedance-2.0-text-to-video. Fast variant: seedance-2.0-fast-text-to-video (supports optional model_params.web_search).
Output quality scales with prompt specificity. Describing camera motion (pan, tilt, zoom, dolly), lighting conditions (time of day, color temperature), subject behavior, and pacing produces significantly better results than generic descriptions.
Image-to-Video
Animates still images with two distinct behaviors:
- 1 image: First-frame animation — the model animates outward from the provided frame
- 2 images: Interpolated transition — smooth motion from first to last frame
{
"model": "seedance-2.0-image-to-video",
"prompt": "Camera slowly pushes in, the still scene comes to life",
"image_urls": ["https://example.com/product.jpg"],
"duration": 5,
"aspect_ratio": "adaptive"
}
The fast variant (seedance-2.0-fast-image-to-video) supports a broader set of input image formats.
Reference-to-Video
Highest-control mode. Accepts multimodal reference inputs via image_urls, video_urls, and audio_urls simultaneously. The model extracts and combines characteristics from all provided references:
- Visual style and composition from images
- Camera movement and motion patterns from video
- Rhythm, pacing, and audio characteristics from audio
{
"model": "seedance-2.0-reference-to-video",
"prompt": "Use video 1's camera movement with audio 1 as background music",
"image_urls": ["https://example.com/character.jpg"],
"video_urls": ["https://example.com/motion-ref.mp4"],
"audio_urls": ["https://example.com/bgm.mp3"],
"duration": 10,
"quality": "720p"
}
Billing note: Input reference video duration is included in cost calculation.
Pricing Model
Credit-based, no subscription. 1 credit = $0.01 USD.
Text-to-video and image-to-video:
| Resolution | Credits/second |
|---|---|
| 480p | 4.63 |
| 720p | 10.00 |
Formula: cost = output_seconds × resolution_rate
Reference-to-video:
Formula: cost = (input_video_seconds + output_seconds) × resolution_rate
Smart duration (duration: -1): Model determines optimal output length. Billing based on actual output duration, not a fixed reservation.
Volume estimates:
| Scenario | Monthly cost |
|---|---|
| 10 vids/day, 5s each, 720p | $150 |
| 20 demos/week, 8s each, 720p | $64 |
Optimization Techniques
- Smart duration (
duration: -1) — Content-aware length selection. Charged for actual output only. - Adaptive aspect ratio (
aspect_ratio: "adaptive") — Model selects optimal ratio for input content, avoiding forced cropping. - Reference video trimming — Since input duration is billed, trim to minimum necessary (5-10s):
ffmpeg -i long.mp4 -t 5 -c copy ref.mp4 - Fast models for iteration — Use
fast-variants during development; switch to standard for production output.
Integration Path Selection
| Path | Best For | Setup Time | Requirement |
|---|---|---|---|
| Playground | Evaluation, prompt testing | 1 min | None |
| ClawHub Skill | Prototyping, creative work | 2 min | Agent environment |
| Direct API | Production, batch processing | 15 min | Developer |
Resources
- API documentation: docs.evolink.ai/en/api-manual/video-series/seedance2.0
- Code examples: github.com/EvoLinkAI/Seedance-2.0-API
- ClawHub skill: clawhub.ai/evolinkai/seedance-2-video-gen
- Full guide: Seedance 2.0 API on EvoLink