Coqui TTS

Coqui TTS is one of the most complete open-source text-to-speech toolkits — a library that lets developers run, train and fine-tune speech synthesis models on their own hardware, with support for dozens of languages and voice cloning from short samples.

Born from Mozilla's TTS research, Coqui packages state-of-the-art architectures behind a usable Python API: generate speech in 1,100+ languages via massively multilingual models, clone a voice from seconds of reference audio (XTTS), stream with low latency, and train custom voices on your own datasets. Because it runs locally, audio never leaves your infrastructure — the privacy and cost profile commercial APIs can't offer at scale.

The software is free and open source. (The company behind it shut down; the models and an active community fork live on — expect community support rather than a vendor.)

Strengths: no per-character API costs, full data privacy, real multilingual breadth, and trainability for custom voices. Weaknesses: developer-oriented setup (Python, GPU for best results), best-model licensing restricts some commercial uses — check the specific model's license — and polish trails paid leaders like ElevenLabs.

Who it's for: developers building voice features at volume, privacy-sensitive projects, researchers, and products needing offline or on-premise speech.