Coqui TTS (Idiap fork)
A maintained fork of the Coqui TTS library with training tools and pre-trained models in over 1,100 languages.
PyPI package coqui-tts with XTTS-v2, VITS and tools to train or fine-tune voices.
This is the Idiap fork of the original Coqui TTS toolkit, which its README calls unmaintained. It suits developers and researchers who want a trainable, multilingual text-to-speech library rather than a finished voice app.
What it is
You install the coqui-tts package with uv or pip after PyTorch, then use the command line, the Python API or the included server. It bundles spectrogram models, end-to-end models such as XTTS and VITS, vocoders and voice conversion, plus dataset analysis tools.
Who it's for
It suits developers building their own voice features and teams training voices for a language. It does not suit anyone who wants a ready studio interface or clear commercial terms for every bundled model.
Key features
Pre-trained models
Fairseq models cover about 1,100 languages.
XTTS-v2
17 languages with streaming under 200 ms latency.
Training and fine-tuning
Recipes and logs with Tensorboard support.
Dataset tools
Utilities to analyse and curate TTS datasets.
Server and API
Command line, Python API and a built-in server.
Voice conversion
Includes OpenVoice and kNN-VC conversion models.
Up and running in 3 steps
Our honest verdict
What's great
- MPL-2.0 library licence with active releases on PyPI under the name coqui-tts.
- Covers training, fine-tuning and dataset curation, not just inference.
- Pre-trained Fairseq models reach about 1,100 languages.
- Command line, Python API and server give several ways to integrate it.
Where it falls short
- Models carry their own licences, including XTTS, so the library licence does not settle commercial use.
- You assemble PyTorch and dependencies yourself; the README gives no VRAM figure.
- It is a library, so there is no voice editor, timeline or hosted API.
- It is a fork of an original repository that its README says is unmaintained.
Pricing & license
Free under MPL-2.0. The only cost is the hardware you run it on; the project shows no paid hosted plan.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Coqui TTS still maintained?
The original repository is described as unmaintained; this Idiap fork publishes the coqui-tts package and shows recent releases.
Can I use XTTS commercially?
The library is MPL-2.0, but each model has its own licence. Read the XTTS model licence before commercial use.
How many languages does it cover?
The README cites pre-trained models in over 1,100 languages, and XTTS-v2 handles 17.
