whatmodel.sh / tts / VoiceCraft 830M (giga830M)
VoiceCraft 830M (giga830M)
↗ share
+ contribute
VoiceCraft 830M (giga830M) is an open-weights text-to-speech model from Puyuan Peng et al. (UT Austin), released March 2024. Its published size is 0.83B parameters. Licence: coqui-public-model-license.
Catalogue identity notes
pyp1/VoiceCraft giga830M.pth (committed 2024-03-28; mirrored as pyp1/VoiceCraft_giga830M), the original 830M GigaSpeech checkpoint evaluated in the paper. Distinct from 830M_TTSEnhanced (voicecraft-830m-tts-enhanced) and the 330M checkpoints. GitHub README: weights under the Coqui Public Model License 1.0.0; code CC BY-NC-SA 4.0.
Model information sources
VoiceCraft paper by the maker (Puyuan Peng et al.). The catalogue checkpoint giga830M is the paper Main 830M GigaSpeech model (official README: "03/28/2024: Model weights for giga330M and giga830M are up"). Covers the VoiceCraft Transformer LM only; the separate Encodec codec is stated in the README as 56M parameters and is not included.
Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.
$ how it compares
The text-to-speech models ranked either side of VoiceCraft 830M (giga830M), out of 148 with a score, then more from Puyuan Peng et al. (UT Austin).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 124 |  VITS (VCTK multi-speaker) | 34±5 | local | ||
| 125 |  Sesame CSM-1B | 34±5 | api $7.00/M chars +local | ||
| 126 | VoiceCraft 830M (giga830M) | 32±4 | local | ||
| 127 |  VALL-E X (Plachtaa open implementation) | 31±4 | local | ||
| 128 |  OuteTTS 0.2 500M | 31±6 | local | ||
| more from Puyuan Peng et al. (UT Austin) | |||||
| 118 | VoiceCraft 830M TTSEnhanced | 37±8 | local | ||
see all text-to-speech models · find the ones that fit your machine
$ what it's good at
other tests3 not in a task · 93% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| VoiceCraft zero-shot TTS listening test (80) · naturalness MOS lab | 4.17 / 5 ✎ | █········█·█ beats 2/2 | 47% |
| VoiceCraft zero-shot TTS set (250) · speaker similarity (WavLM-TDCNN) lab | 55.0% ✎ | █····█·····█ beats 2/2 | 24% |
| VoiceCraft zero-shot TTS set (LibriTTS + YouTube, 250) · Whisper medium.en WER lab | 4.5% ✎ | █······█···█ beats 1/2 | 23% |
$ where to run it
Other ways to run it
- Run it as published · needs ~4.53 GB · from github.com/jasonppy
git clone https://github.com/jasonppy/VoiceCraft && cd VoiceCraft && conda create -n voicecraft python=3.9.16 && conda activate voicecraft && pip install -e git+https://github.com/facebookresearch/audiocraft.git@c5157b5bf14bf83449c17ea1eeb66c19fb4bc7f0#egg=audiocraft && pip install xformers==0.0.22 && pip install torchaudio==2.0.2 torch==2.0.1 && apt-get install ffmpeg && apt-get install espeak-ng && pip install tensorboard==2.16.2 && pip install phonemizer==3.2.1 && pip install datasets==2.16.0 && pip install torchmetrics==0.11.1 && pip install huggingface_hub==0.22.2 && conda install -c conda-forge montreal-forced-aligner=2.2.17 openfst=1.8.2 kaldi=5.5.1068 && mfa model download dictionary english_us_arpa && mfa model download acoustic english_us_arpa && python3 tts_demo.py -m giga830M
Weights and documentation: huggingface.co/pyp1 ↗
$ how the score was worked out · VoiceCraft: Zero-Shot Speech… 93%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 VoiceCraft: Zero-Shot Speech Editing and Text-to-Speech in the Wild (arXiv:2403.16973v3) · lab-reported | 3 | 93% |
| prior · pull toward the middle | 7% | |
| weighted average | = 32.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json