Skip to Content

Video Generation

Create AI-generated videos from text prompts.

Stable

Overview

elizaOS Cloud provides video generation through Fal.ai, supporting multiple models:

Quick Start

Dashboard

Navigate to Dashboard → Videos for the visual interface.

API

curl -X POST "https://cloud.milady.ai/api/v1/generate-video" \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "prompt": "A butterfly emerging from a cocoon in slow motion", "model": "fal-ai/veo3" }'

Response

{ "success": true, "video": { "url": "https://your-storage.vercel-storage.com/videos/abc123.mp4" }, "model": "fal-ai/veo3" }

Available Models

ModelProviderQualitySpeed
`fal-ai/veo3`GoogleHighestSlow
`fal-ai/veo3/fast`GoogleHighFast
`fal-ai/kling-video/v2.1/master/text-to-video`KlingHighestSlow
`fal-ai/kling-video/v2.1/pro/text-to-video`KlingHighMedium
`fal-ai/kling-video/v2.1/standard/text-to-video`KlingGoodFast
`fal-ai/minimax/hailuo-02/pro/text-to-video`MiniMaxHighMedium
`fal-ai/minimax/hailuo-02/standard/text-to-video`MiniMaxGoodFast

Request Parameters

ParameterTypeRequiredDescription
`prompt`stringText description of the video
`model`stringModel to use (default: `fal-ai/veo3`)

Prompt Tips

Camera Movements

Include camera direction for dynamic videos:

Scene Description

Be specific about motion and timing:

{ "prompt": "A golden retriever running through autumn leaves in a park, slow motion, cinematic lighting, tracking shot following the dog" }

Style Modifiers

Best Practices

  1. Be specific about motion: Describe what moves and how
  2. Include timing cues: Slow motion, timelapse, etc.
  3. Specify camera work: Pan, zoom, tracking shots
  4. Set the mood: Cinematic, documentary, etc.

Video generation typically takes 30-180 seconds depending on model and complexity.

Error Handling

CodeErrorSolution
400Invalid promptEnsure prompt is non-empty
400Invalid modelUse a supported model
402Insufficient creditsAdd credits to your account
429Rate limitedWait and retry
500Generation failedRetry or try a different model

Next Steps