Avatars

AI avatar generation pipeline - create animated talking avatars from photos.

Pipeline

Photo → Style Transfer → TTS Audio → Animation → Talking Video

Setup

# Install
uv sync

# Set Replicate API token
export REPLICATE_API_TOKEN=your_token_here

Get a token at: https://replicate.com/account/api-tokens

Usage

Full Pipeline

Generate a complete talking avatar video:

# Using an executive's photo
avatars generate steve-butcher "Welcome to the Q3 update. We've made significant progress on the metro project."

# With options
avatars generate steve-butcher "Hello everyone" --style pixar --voice male_british -o output.mp4

# Skip style transfer (use original photo)
avatars generate steve-butcher "Hello" --skip-style

Individual Steps

# List available executives
avatars list

# Just stylize a photo
avatars stylize steve-butcher --style illustration

# Just generate speech
avatars speak "Hello world" --voice male_british -o speech.wav

# Just animate (image + audio → video)
avatars animate photo.jpg audio.wav -o video.mp4

Style Presets

• illustration - Clean corporate illustration style
• pixar - 3D Pixar-style character
• cartoon - Flat cartoon/vector style

Voice Presets

• male_british - British male voice
• male_american - American male voice
• female_british - British female voice
• female_american - American female voice

Adding Executives

Create a folder in executives/ with:

executives/
  john-smith/
    metadata.json
    photo.jpg

metadata.json:

{
  "id": "john-smith",
  "name": "John Smith",
  "title": "CEO",
  "company": "Acme Corp",
  "photos": {
    "primary": "photo.jpg"
  }
}

Models Used

• Style Transfer: Stable Diffusion XL (img2img)
• TTS: Bark (Suno AI)
• Animation: LivePortrait / SadTalker

All models run via Replicate API by default. Local execution support planned.