Offline Electron AI Avatar Integration
Budget: $100 – $300 USD
My desktop app is built with Electron and I now want to let users generate and converse with a personalised AI avatar entirely on-device. The new module should:
• Accept a few static images and a short voice sample, then create a talking head that mirrors natural face movement and lip-syncs.
• Clone the user’s voice so replies sound like them.
• Run inference offline; once assets are cached, no cloud call should be required for real-time dialogue.
You’ll wire this into the existing Electron UI (React on the front end) by adding a “Create Avatar” flow, background workers for the ML models, and a simple conversation panel. Feel free to choose proven open-source tools—e.g. TensorFlow.js, MediaPipe, Vosk, or similar—so long as licensing is compatible with commercial distribution and performance stays under a few hundred ms per frame on a modern laptop.
Deliverables:
1. Integrated Electron module (TypeScript preferred) with build scripts.
2. Model weights and conversion steps documented for offline use.
3. Brief README covering installation, asset requirements, and how to extend or retrain.
I’ll give you repo access and sample media as soon as we start. Let me know which libraries you’d use and how you’ll keep the footprint manageable for an offline desktop environment.
• Accept a few static images and a short voice sample, then create a talking head that mirrors natural face movement and lip-syncs.
• Clone the user’s voice so replies sound like them.
• Run inference offline; once assets are cached, no cloud call should be required for real-time dialogue.
You’ll wire this into the existing Electron UI (React on the front end) by adding a “Create Avatar” flow, background workers for the ML models, and a simple conversation panel. Feel free to choose proven open-source tools—e.g. TensorFlow.js, MediaPipe, Vosk, or similar—so long as licensing is compatible with commercial distribution and performance stays under a few hundred ms per frame on a modern laptop.
Deliverables:
1. Integrated Electron module (TypeScript preferred) with build scripts.
2. Model weights and conversion steps documented for offline use.
3. Brief README covering installation, asset requirements, and how to extend or retrain.
I’ll give you repo access and sample media as soon as we start. Let me know which libraries you’d use and how you’ll keep the footprint manageable for an offline desktop environment.