YggSaynaAI/sayna
↗ GitHubSayna is a unified Voice Layer for AI Agents with a seemless integration to an existing agentic frameworks
165
Stars
27
Forks
2
Watchers
0
Open Issues
Safety Rating A
No hardcoded secrets, malicious code patterns, suspicious dependencies, or prompt injection attempts were identified. API keys are referenced only via environment variables and configuration files, with no credentials embedded in source. The repository appears to be a legitimate open source Rust server project.
ℹAI-assisted review, not a professional security audit.
AI Analysis
Sayna is a high-performance real-time voice processing server written in Rust that provides a unified Speech-to-Text (STT) and Text-to-Speech (TTS) layer for AI agents. It exposes WebSocket and REST APIs, supports multiple providers (Deepgram, ElevenLabs, Google Cloud, Microsoft Azure), integrates with LiveKit for WebRTC audio streaming, and optionally includes Silero-VAD turn detection and DeepFilterNet noise suppression. It is designed to be embedded as a voice layer into existing agentic frameworks.
Use Cases
- Adding real-time voice I/O to AI agent pipelines
- Streaming speech-to-text transcription in WebRTC/LiveKit rooms
- Converting LLM text responses to speech via multiple TTS providers
- Building voice-enabled AI assistants with VAD and turn detection
- Deploying a self-hosted voice API server with provider-agnostic abstraction
Tags
Project Connections
ComfyUI-Qwen-TTS
→Both projects provide TTS/voice synthesis capabilities for AI workflows, but Sayna is a standalone real-time voice server supporting multiple providers and STT, whereas ComfyUI-Qwen-TTS is a ComfyUI plugin focused on Qwen TTS model integration.
murmure
→Both handle speech-to-text processing, but murmure is a local-first desktop STT app using an on-device model, while Sayna is a server-side voice layer supporting cloud STT/TTS providers with WebSocket streaming.
CoPaw
→CoPaw is a multi-agent AI assistant framework that lists voice as a feature; Sayna could serve as its dedicated voice processing backend, providing the unified STT/TTS API layer that CoPaw's voice channel would consume.
zeroleaks
→Sayna exposes WebSocket and REST endpoints for AI agent voice pipelines; zeroleaks could be used to audit those endpoints for prompt injection or system prompt extraction vulnerabilities in the agent flows behind Sayna.