Izwi

Izwi: On-Device Voice AI for Production

Izwi is an open-source, production-ready on-device voice AI engine built for real-time applications. It allows developers to run voice AI locally with sub-100ms latency, eliminating cloud dependencies, API costs, and privacy concerns. The platform features a Rust-native engine optimized for production workloads, supporting both Apple Silicon and NVIDIA CUDA for GPU acceleration.

Key Features

• Complete Voice AI Stack: Includes text-to-speech, speech recognition with word-level timestamps, voice cloning from seconds of audio, voice design from text descriptions, conversational AI, and speaker diarization.
• 100% Private Local Processing: Data never leaves the machine, requiring no internet connection, API keys, or usage limits.
• Izwi Desktop Playground: A native desktop application for macOS, Windows, and Linux featuring a visual interface for prototyping, real-time conversations, and drag-and-drop audio transcription.
• Server and API: A production-ready HTTP API server with OpenAI-compatible endpoints and WebSocket support for streaming, acting as a drop-in replacement for cloud APIs.
• High Performance: Delivers sub-100ms Time to First Token latency using a Rust-native inference engine.

Pricing

Izwi is open-source and free forever under the Apache 2.0 License.

Getting Started

Website: https://izwiai.com/

Izwi provides developers with a complete, private, and high-performance voice AI infrastructure that runs entirely on-device without cloud costs.