11 models · updated 2026-09-02
Text to speech and music models: what your GPU runs
Local text to speech and music models are small by current standards. The interesting question is rarely whether they fit, but whether they generate faster than the audio plays.
Results
| Model | Denoiser | Verdict | Arrangement | Peak VRAM | Speed |
|---|---|---|---|---|---|
| Loading… | |||||
Every model in this catalog
Download sizes are the full pipeline at the precision it is published in.
Kokoro 82M
0.082B denoiser · 0.082B total · 0.3 GB to download
XTTS v2
0.467B denoiser · 0.467B total · 1.7 GB to download
Whisper Large v3
3.087B denoiser · 3.087B total · 8.6 GB to download
Whisper Small
0.242B denoiser · 0.242B total · 0.9 GB to download
MusicGen Medium
1.5B denoiser · 1.5B total · 5.6 GB to download
Chatterbox
2.67B denoiser · 2.67B total · 9.9 GB to download
F5-TTS
0.337B denoiser · 0.337B total · 1.3 GB to download
VibeVoice Realtime 0.5B
1.018B denoiser · 1.018B total · 1.9 GB to download
MusicGen Small
0.591B denoiser · 0.591B total · 2.2 GB to download
MusicGen Large
3.3B denoiser · 3.3B total · 12.3 GB to download
Stable Audio Open
1.057B denoiser · 1.323B total · 4.9 GB to download