Fay

Why we included this project

Fay suits teams that want a digital human doing real work against their own systems rather than just chatting. The framework pairs the avatar format you pick, 2.5D, 3D, PC, mobile, or web, with an OpenAI-compatible or DeepSeek model, and lets you swap in your preferred speech recognition and text-to-speech engines. It also carries genuine agent behavior: autonomous tool calls, MCP tool management, scheduled proactive dialogue, and wake and interrupt handling, all steered from a configuration center and a browser-based admin page. Because the whole pipeline can run offline, it works for virtual anchors, virtual teachers, and voice-driven kiosks that cannot depend on a live connection. Separate interfaces for text, voice, avatar driving, and control make it an approachable reference when you are wiring an avatar front end to your own backend.