coqui-tts

What it is and what it does

Coqui TTS is a deep learning library for converting text into natural-sounding speech. It bundles multiple neural architectures (Tacotron2, Glow-TTS, VITS, XTTS, and others) with pretrained weights across many languages, plus vocoders to convert spectrograms to audio. You can use it off-the-shelf for inference, fine-tune existing models on your own data, or train new models from scratch. It also supports voice conversion (changing a speaker's identity while preserving content) and voice cloning with minimal reference audio.

The library is designed for both research and production use. It provides command-line tools and a Python API, with utilities for dataset curation and analysis. The main constraint is that external PyTorch installation is required, and the full dependency stack (transformers, librosa, scipy, numba, and others) is substantial. Training new models is computationally expensive; inference can run on CPU but is much faster on GPU.

Use it for:

• Generate speech from text in multiple languages using pretrained models without training
• Fine-tune an existing TTS model on your own voice or dataset to customize output
• Build a voice cloning system that generates speech in a target speaker's voice from a short audio sample
• Convert one speaker's voice to another while preserving the linguistic content
• Train a multilingual or multi-speaker TTS model from scratch on custom data
• Analyze and curate TTS training datasets using built-in dataset analysis tools

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Coqui TTS synthesizes speech from text using deep learning models, supporting over 1100 languages with pretrained weights and tools for training and fine-tuning custom models.

Yes, if you need multilingual text-to-speech synthesis or voice conversion. The library is actively maintained, has no known vulnerabilities, and offers both easy inference and advanced training capabilities. Install it if you want pretrained models out-of-the-box or plan to fine-tune. The MPL-2.0 copyleft license is permissive for unmodified use in closed-source work but requires sharing modifications. Be aware that external PyTorch installation is required and the dependency footprint is large.

Install

coqui-tts on PyPI

pip install coqui-tts

uv add coqui-tts

poetry add coqui-tts

Installing coqui-tts

Before you install

Low friction installation via wheel distribution. Active maintenance with last commit 2026-06-10. Requires Python 3.10–3.14. The 21 runtime dependencies include heavy scientific stacks (transformers, librosa, scipy, numba) typical of ML libraries.

License in practice

MPL-2.0 is copyleft: derivative works and modifications must be distributed under the same license. If you modify the library itself, you must share those changes. Using it unmodified in closed-source applications is permitted.

Quickstart

pip install coqui-tts

from TTS.api import TTS
tts = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", gpu=False)
tts.tts_to_file(text="Hello world", file_path="output.wav")

Requires Python 3.10–3.14. External PyTorch installation required (not bundled since 0.27.4). GPU support optional but recommended for inference speed.

Verify before relying

• Whether the 1100 languages claim refers to Fairseq models or all available models in the library
• Inference latency and memory requirements for typical use cases
• Whether voice cloning (mentioned in 0.27.0 news) requires additional setup or training data
• Exact PyTorch version requirements and compatibility with PyTorch 2.2+

Package facts

Evidence: coqui_tts-0.27.5-py3-none-any.whl

Tags

More Software Development packages