MCP API Integration Guide · Video
MCP API — Video Creation
Guide to creating AI video clips and composing them into final videos via MCP — focused on create_clip and compose_video.
Overview
This section covers creating AI video clips and composing them into a final video via MCP. The video pipeline is asynchronous and uses create_clip and compose_video tools sequentially.
create_clip
Accepts text, image, or both as input and generates an AI video clip. Returns a taskId; poll with get_job_status.
Modes
- text-to-video — prompt only. Default model:
veo3-fast - image-to-video — single image to video. Default model:
kling-v3-pro - image-text-to-video — image + motion prompt. Default model:
kling-v3-pro
Supported Models
| Model | Duration | Features |
|---|---|---|
veo3-fast | 5-8s | T2V fast, auto audio |
veo3 | 5-8s | T2V highest quality, auto audio |
kling-v3-pro | 3-15s | I2V/IT2V, characterReferences + endImage |
kling-o3 | 3-15s | I2V/IT2V, same features |
kling-v2.6-pro | 5 or 10s | endImage supported |
ltx-2.3 | 6-20s (even) | endImage, fast, no audio |
seedance-2.0 | 4-15s | All modes, auto audio |
seedance-2.0-fast | 4-15s | Fast variant |
Pricing
- 5s — 2,500 coins
- 10s — 5,000 coins
- 15s — 7,500 coins
Failed jobs are automatically refunded.
Character Consistency
Use characterReferences to keep the same person across multiple scenes (kling-v3-pro, kling-o3 only). Pass frontal_image_url or assetId.
Scene Continuity (multi-scene chaining)
To preserve motion/camera from the previous scene, pass previousClipTaskId, referenceVideoUrl, continueMotion, preserveCamera, preserveSceneLayout.
compose_video
Composes completed clips into a final video using Remotion with transitions, color grading, branding, and BGM.
Key Settings
- format — 16:9 / 9:16 / 4:3
- transition — crossfade, fade, slide-*, wipe-*, clock-wipe, flip, none
- colorGrading — none / warm / cool / cinematic / vintage
- vignette, filmGrain — VFX toggles
- brand — brand name + colors (used in captions/outro)
Automatic BGM Mixing
If a BGM node is connected to the compose node in your workflow graph, it is automatically mixed during composition (crossfade loop + fade-out + ducking by default). See the "Audio / BGM" section for details.
Pricing
500 coins per composition.
Example: 30-second Ad Video
upload_assetorgenerate_image→ get a character assetIdcreate_workflow→ get workflowIdcreate_clip× 3-6 times (one per scene)get_job_statuspolling until all clips are done- (Optional)
create_bgmfor background music compose_videofor final compositionget_job_statusuntil composition done → outputUrl