#1
Run LLMs, image, and audio models locally with an OpenAI-compatible API, optional GPU acceleration, and a built-in web UI for managing and testing models.
48.7k stars4.4k forksMITlast commit Actively maintained
3 self-hosted Speechmatics alternatives
Speechmatics is a cloud-based automatic speech recognition and text-to-speech service offering real-time and batch transcription via APIs. It supports 55+ languages, domain-customizable models, low-latency streaming, and enterprise privacy/compliance deployment options.
Alternatives to Speechmatics
#2
Self-hosted, OpenAI API-compatible server for streaming transcription, translation, and speech generation using faster-whisper and TTS engines like Piper and Kokoro.
3.6k stars444 forksMITlast commit Actively maintained
#3
Self-hosted voice assistant platform for ESP32 devices with on-device wake-word and command recognition, Home Assistant integration, and an optional inference server for STT/TTS/LLM.
3.1k stars128 forksApache-2.0last commit Actively maintained
What replacing Speechmatics actually involves
Every option on this page is open source and free to run on your own hardware, so you own the data and there is no subscription to cancel. 3 of 3 shipped a commit in the last six months. Licences in this list: MIT, Apache-2.0. In exchange you take on hosting, backups and updates yourself.
Browse everything in Model Serving & Inference.
