Building a Voice AI Platform with 28 Modules in Python

Short summary

Omni-VRAM is an open-source Python voice AI platform with 28 modules supporting multiple speech recognition backends (Whisper, ONNX, TensorRT), real-time streaming under 200ms, speaker diarization, and emotion recognition. It includes Chinese language support, meeting auto-summarization, and text-to-speech synthesis. Multiple deployment APIs (REST, WebSocket, gRPC) and Docker support enable flexible integration for GPU and CPU environments.

• •28 modular components for speech recognition, synthesis, diarization, and emotion detection
• •Sub-200ms latency with multiple backend options and multilingual support
• •Flexible deployment via REST/WebSocket/gRPC APIs with GPU and CPU Docker containers

Generated with AI, which can make mistakes.