Speech Swift

Speech Swift, published under the Soniqo name, is an open source toolkit for running speech AI directly on a user's own device. It handles diarized transcription, zero shot voice cloning and long form speech synthesis with no cloud API calls, no per minute fees and no audio leaving the device. It targets Apple Silicon first, with support for Android, Windows and embedded Linux, and is released under the Apache 2.0 license. The audience is developers building products with voice features. The stack bundles more than thirty models, including speech to text options such as Qwen3 ASR, Whisper Large v3 Turbo, Parakeet TDT and a diarization model, plus speech generation and voice agent components. Use cases are grouped into conversational voice agents, including full duplex speech to speech with tool calling over MCP, transcription for live captions or batch archives, and speech generation for audiobooks and multi speaker podcasts. Installation is through Homebrew for Apple platforms or a Gradle dependency for Android. Soniqo also offers Stenograf, an app that records calls and meetings, transcribes them on the user's computer, and answers questions about them using quoted passages. No bot joins the call and nothing is sent to a transcription service. The library is free to use under its license. The site publishes no pricing for Stenograf.