Tortoise TTS

Free plan APILinuxMacSelf-hostedWindows

What Tortoise TTS is

Tortoise TTS is free, Apache 2.0-licensed text-to-speech software focused on multi-voice output and realistic prosody and intonation. It runs on Linux, macOS and Windows, supports self-hosting and provides API access. The repository contains inference code, and its Python API accepts reference audio plus a selected preset to generate speech. Speaker reference clips guide pitch, tone and speaking speed. Custom voices require at least three roughly 10-second WAV clips recorded at 22,050 Hz. For long text, the read and read_fast scripts split files into speech clips and can combine them into one file. A socket streaming server listens on port 5000. A hosted demonstration is available through Hugging Face Spaces, although CPU-only Spaces do not work for it. Local operation requires an NVIDIA GPU; separate Apple Silicon instructions cover macOS 13 or later with M1 or M2 chips. The project acknowledges Hugging Face for its GPT model, generation API and model-weight hosting.

Who it suits

It suits developers and creators who need local, self-hosted or API speech generation with reference-guided voices. Users should have supported hardware for local inference or use the hosted demonstration.

What it does well

Apache 2.0 open-source license
Reference clips guide pitch, tone and speed
Custom voices use WAV reference clips
Scripts handle long text files
Python API and socket streaming server included

What to know first

Local use requires an NVIDIA GPU
CPU-only Hugging Face Spaces do not work
Custom voices require at least three clips
Apple Silicon setup targets macOS 13 or later

Plans

Plan	Billed and limits	Price
Open-source software	Apache 2.0 licensed code · run inference locally or use a hosted demo	Free

As github.com lists them, read 12 Oct 2026. The maker's own page has the final word.

Verdict

Tortoise TTS offers open-source speech generation with multi-voice controls, reference audio and tools for long text. Hardware requirements matter for local use, while the repository also provides API, streaming and hosted-demo routes.

Facts from Tortoise TTS's own pages

API access: Yes · github.com, 20 Sept 2026
Custom voices: Users can add a voice by providing at least three roughly 10-second WAV clips at 22,050 Hz. · github.com, 12 Oct 2026
Dependencies: The project acknowledges Hugging Face for the GPT model and generation API and for hosting model weights. · github.com, 12 Oct 2026
Hardware: The README says local use requires an NVIDIA GPU and describes a separate Apple Silicon setup for macOS 13 or later with M1 or M2 chips. · github.com, 12 Oct 2026
Hosted demo: A live demo is hosted on Hugging Face Spaces, and the README says CPU-only Spaces do not work for it. · github.com, 12 Oct 2026
Inference: The repository contains the code needed to run Tortoise TTS in inference mode. · github.com, 12 Oct 2026
License: Tortoise TTS is licensed under Apache 2.0. · github.com, 12 Oct 2026
Long text: The read and read_fast scripts split text files into speech clips and can combine the clips into one file. · github.com, 12 Oct 2026
Maker: The author identifies himself as James Betker and says his employer was not involved in the project’s development. · github.com, 12 Oct 2026
Platforms: windows,macos,api,self_hosted · github.com, 20 Sept 2026
Programmatic use: The Python API accepts reference audio and generates speech using a selected preset. · github.com, 12 Oct 2026
Purpose: Tortoise is a text-to-speech program focused on multi-voice capability and realistic prosody and intonation. · github.com, 12 Oct 2026
Streaming: The repository includes a socket streaming server that listens on port 5000. · github.com, 12 Oct 2026
Voice cloning: Yes · github.com, 20 Sept 2026
Voice references: The model uses speaker reference clips to guide voice qualities such as pitch, tone, and speaking speed. · github.com, 12 Oct 2026

Where Tortoise TTS stands on Freedom251

Text-to-speech tools#58 of 232
Text-to-Speech Software#90 of 223