Tortoise TTS
neonbjb/tortoise-ttsReviewed by AI
Checked by RepoPicks Radar on
What is Tortoise TTS?
Tortoise is a text-to-speech program made for natural-sounding intonation and many voices. You give it a sentence or a whole text file and it writes out spoken audio; give it a few recorded clips of a voice and it will read in that voice. It is used from the command line or from Python, with scripts for one phrase, for long texts and for streaming. Running it on your machine means a conda environment, Python 3.9 and an NVIDIA card; an Apple Silicon path exists but runs on the processor. A free demo page exists on Hugging Face.
Tortoise TTS is free and open source (Apache-2.0) and can replace ElevenLabs, Murf, and Speechify.
How easy is it to install Tortoise TTS?
A · download and open · Web
- Windows D
- macOS D
- Linux D
- Web A
The quick way is the live demo page on Hugging Face linked in the README, which needs nothing installed. Every local route is terminal work: "If you want to use this on your own computer, you must have an NVIDIA GPU", with miniconda, a Python 3.9 environment and a matching PyTorch build on Windows and Linux, or a Docker image; on Apple Silicon you install a nightly PyTorch, DeepSpeed is disabled and you may have to set `PYTORCH_ENABLE_MPS_FALLBACK=1`.
Tortoise TTS on GitHub
- Stars
- 14,894on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +1stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux, Web
- You need to install
- Nothing extra
- To build from source
- Python, Docker, FFmpeg
- Disk space
- 150 MB–450 MB · estimate
- Accounts or keys
- None
- Cost
- Free · Free and Apache 2.0, model weights hosted by Hugging Face. Duplicating the demo page with a graphics card, as the README suggests to skip the queue, is paid on Hugging Face.
- License
- Apache-2.0 (Permissive)
- Can replace
- ElevenLabs, Murf, Speechify
- Last change on GitHub
- Nov 19, 2024
Tortoise TTS risk signals (5)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): transformers: GHSA-29pf-2h5f-8g72, GHSA-37mw-44qp-f5jm, GHSA-37q5-v5qm-c9v8; nbconvert: GHSA-9754-6rhw-gj3h, GHSA-9jmq-rx5f-8jwq, GHSA-xm59-rqc7-hhvf; tornado: GHSA-3hv7-mjh2-fv65, GHSA-3x9g-8vmp-wqvf, GHSA-753j-mpmx-qq6g; pydantic: GHSA-mr82-8j83-vxmv, PYSEC-2026-1812; deepspeed: GHSA-8cp5-3rf8-8gfh, PYSEC-2024-109
automatic check
- Attentionvoice cloning impersonation risk
The API takes `reference_clips` of a real voice (`tts.tts_with_preset("your text here", voice_samples=reference_clips)`) and the repository ships a `voice_customization_guide.md`, so anyone's voice can be copied from recordings; there is no disclaimer about it in the README.
AI review
- Attentionunmaintained
The last commit is 2024-11-19, 719 days before this check, with 349 issues open and no releases ever published.
AI review
- Infovery old dependency pins
requirements.txt pins `tornado==4.2`, `nbconvert==5.3.1`, `pydantic==1.9.1`, `transformers==4.31.0` and `deepspeed==0.8.3`, and the README itself warns that without conda "you will spend a lot of time chasing dependency problems".
AI review
- Infocontradictory speed claims
The README says both "this model is insanely slow ... expect to generate a medium sized sentence every 2 minutes" and "not so slow anymore now we can get a **0.25-0.3 RTF** on 4GB vram", with no explanation of which setup gives which.
AI review
Evidence is shown in its original language.
What Radar could not check about Tortoise TTS
- How long one sentence really takes on a current graphics card, and on Apple Silicon
- Whether the conda and pip installs still complete with these pinned versions
- How much disk space the model files take once downloaded from Hugging Face
- Whether the community-run demo page stays online and what it does with the audio you upload
- Whether the known problems in the pinned packages found by the checks matter for local use