seedance-ai-video（种子 AI 视频）

Seedance — AI Video Generation by ByteDance

Generate AI videos with synchronized audio using ByteDance's Seedance 1.5 Pro — featuring native audio-visual joint generation, cinematic camera control, multi-language lip-sync, and diverse sound effects.

Seedance excels at creating cinematic short clips with realistic motion, facial expressions, spatial audio, and complex camera movements.

Data usage note: This skill sends text prompts and image URLs to the Atlas Cloud API (api.atlascloud.ai) for video generation. No data is stored locally beyond the downloaded output files. API usage incurs charges based on the model selected.

Key Capabilities

• Text-to-Video — Generate video clips from text descriptions with synchronized audio
• Image-to-Video — Animate still images into dynamic video with motion and audio
• Native Audio Generation — Dialogue, sound effects, and music generated jointly with video (not post-processed)
• Multi-Language Lip-Sync — English, Chinese (including dialects), Japanese, Korean, Portuguese, Spanish, Indonesian
• Cinematic Camera Control — Dolly-in, snap zoom, first-person POV, tripod lock, crane shots
• Multiple Styles — Realistic, anime, 2D animation, steampunk, ink-wash, and more
• Resolution — Up to 720p (Pro), 480p available
• Duration — 5-12 seconds per clip

Setup

1. Sign up at https://www.atlascloud.ai
2. Console → API Keys → Create new key
3. Set env: export ATLASCLOUD_API_KEY="your-key"

Script Usage

This skill includes a Python script for video generation. Zero external dependencies required.

List available video models

python scripts/generate_video.py list-models

Generate a video

python scripts/generate_video.py generate \
  --model "bytedance/seedance-v1.5-pro/text-to-video" \
  --prompt "Your prompt" \
  --output ./output

Image-to-video

python scripts/generate_video.py generate \
  --model "bytedance/seedance-v1.5-pro/image-to-video" \
  --image "https://example.com/photo.jpg" \
  --prompt "Animate" \
  --output ./output

Run python scripts/generate_video.py generate --help for all options.

Pricing

Pro tier delivers higher quality with more detail and coherence. Fast tier is ~12x cheaper and suitable for drafts and iteration.

Available Models

Text-to-Video

Image-to-Video

Parameters

Text-to-Video

Image-to-Video

Same as text-to-video, plus:

Workflow: Submit → Poll → Download

Text-to-Video Example

# Step 1: Submit
curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateVideo" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance/seedance-v1.5-pro/text-to-video",
    "prompt": "A woman walks through a sunlit bamboo forest, camera slowly dollying forward. Birds chirping in the background, gentle wind rustling leaves.",
    "aspect_ratio": "16:9",
    "duration": 5,
    "resolution": "720p",
    "generate_audio": true
  }'
# Returns: { "code": 200, "data": { "id": "prediction-id" } }

# Step 2: Poll (every 5 seconds until "completed" or "succeeded")
curl -s "https://api.atlascloud.ai/api/v1/model/prediction/{prediction-id}" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY"
# Returns: { "code": 200, "data": { "status": "completed", "outputs": ["https://...video-url..."] } }

# Step 3: Download
curl -o output.mp4 "VIDEO_URL_FROM_OUTPUTS"

Image-to-Video Example

curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateVideo" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance/seedance-v1.5-pro/image-to-video",
    "image": "https://example.com/portrait.jpg",
    "prompt": "The person slowly turns their head and smiles, camera gently zooms in",
    "aspect_ratio": "9:16",
    "duration": 5,
    "generate_audio": true
  }'

Fast Model Example (Quick Draft)

curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateVideo" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance/seedance-v1.5-pro/text-to-video-fast",
    "prompt": "Ocean waves crashing on a rocky shore at sunset, seagulls flying overhead",
    "aspect_ratio": "16:9",
    "duration": 5,
    "generate_audio": true
  }'

Polling Logic

• processing / starting / running → wait 5s, retry (Pro takes ~30-60s, Fast takes ~10-20s)
• completed / succeeded → done, get URL from data.outputs[]
• failed → error, read data.error

Atlas Cloud MCP Tools (if available)

If the Atlas Cloud MCP server is configured, use built-in tools:

atlas_quick_generate(model_keyword="seedance 1.5", type="Video", prompt="...")
atlas_generate_video(model="bytedance/seedance-v1.5-pro/text-to-video", params={...})
atlas_get_prediction(prediction_id="...")

Implementation Guide

1. Determine task type:

- Text-to-video: user describes a scene/action in text - Image-to-video: user provides an image to animate

1. Choose model:

- Pro for final output, client-facing content, or quality-critical use - Fast for quick iteration, drafts, or budget-conscious use

1. Extract parameters:

- Prompt: describe scene, action, camera movement, and audio cues - Aspect ratio: infer from context (social reel→9:16, YouTube→16:9, square→1:1, cinematic→21:9) - Duration: default 5s, up to 12s for longer scenes - Audio: enabled by default; disable with generate_audio: false if user only wants silent video - Camera: set camera_fixed: true for static/tripod shots

1. Execute: POST to generateVideo API → poll result → download MP4

1. Present result: show file path, offer to play

Prompt Tips

Seedance produces best results when prompts describe both visual and audio elements:

• Scene + Action: "A chef flips a pancake in a busy kitchen, sizzling sounds and clattering pans"
• Camera direction: "Camera slowly pans left to reveal...", "Close-up tracking shot of...", "First-person POV walking through..."
• Audio cues: Include sound descriptions — "birds chirping", "rain on window", "jazz music playing softly"
• Dialogue: For talking videos, include speech in quotes — "The narrator says: 'Welcome to our city'"
• Style: "cinematic", "anime style", "documentary", "slow motion", "timelapse"
• Lip-sync: For multi-language dialogue, specify the language — "A woman speaking Japanese says: 'こんにちは'"

Coming Soon: Seedance 2.0

Seedance 2.0 is ByteDance's next-generation unified multimodal video generation system, currently in preview. When available on Atlas Cloud, this skill will be upgraded with:

• Higher resolution — Expected support for 1080p and above
• Longer duration — Extended video length beyond 12 seconds
• Multimodal references — Video-to-video, audio-guided generation
• Director-level control — Fine-grained manipulation of performance, lighting, shadow, and camera
• Enhanced motion stability — Improved realism and coherence across longer clips

The API workflow and parameter structure are expected to remain compatible. Model IDs will be updated when Seedance 2.0 becomes available — no configuration changes needed on your end.