fal-lip-sync

Create talking head videos, sync lips to audio, and animate portraits using fal.ai models.

Scripts

| Script | Purpose | |--------|---------| | talking-head.sh | Generate a talking head video from an image + audio/text | | lip-sync.sh | Sync lips in an existing video to new audio |

Usage

Talking Head (Image + Audio → Video)

./scripts/talking-head.sh --image-url "https://example.com/portrait.jpg" --audio-url "https://example.com/speech.mp3" --model veed/fabric-1.0

Talking Head (Image + Text → Video with auto TTS)

./scripts/talking-head.sh --image-url "https://example.com/portrait.jpg" --text "Hello, welcome to our presentation" --model fal-ai/creatify/aurora

Lip Sync (Video + Audio → Synced Video)

./scripts/lip-sync.sh --video-url "https://example.com/video.mp4" --audio-url "https://example.com/new-speech.mp3"

Arguments

talking-head.sh

| Argument | Description | Required | |----------|-------------|----------| | --image-url | URL of portrait/face image | Yes | | --audio-url | URL of audio to sync | Yes (or --text) | | --text | Text to speak (auto TTS) | Yes (or --audio-url) | | --model / -m | Model endpoint | No (default: veed/fabric-1.0) | | --tts-model | TTS model for --text mode | No (default: fal-ai/minimax/speech-2.6-turbo) | | --wait / -w | Wait for completion | No (default: true) | | --async / -a | Return request ID immediately | No |

lip-sync.sh

| Argument | Description | Required | |----------|-------------|----------| | --video-url | URL of video to lip sync | Yes | | --audio-url | URL of audio to sync to | Yes | | --model / -m | Model endpoint | No (default: fal-ai/sync-lipsync/v2) |

Finding Models

To discover the best and latest lip sync and talking head models, use the search API:

# Search for talking head models
bash /mnt/skills/user/fal-generate/scripts/search-models.sh --query "talking head"

# Search for lip sync models
bash /mnt/skills/user/fal-generate/scripts/search-models.sh --query "lip sync"

# Search for live portrait / expression transfer
bash /mnt/skills/user/fal-generate/scripts/search-models.sh --query "live portrait"

Or use the search_models MCP tool with relevant keywords like "lip sync", "talking head", "avatar".

Output Format

{
  "video": {
    "url": "https://fal.media/files/...",
    "content_type": "video/mp4"
  }
}

Present the video URL directly to the user.