MOSS TTS Nano 100M
last releaseMay 6, 2026
powered byMOSS-TTS-Nano 100M
goblin vibe check:
good if you need local realtime tts on ordinary hardware instead of a giant voice stack
tiny multilingual text-to-speech model built for cpu-first realtime voice generation and cloning. useful when you need a lightweight local speech stack for prototypes, dialogue previews, or tools that cannot afford a heavy gpu tts setup.
speed
real-time
0.1b parameter tiny speech modelrealtime generation on 4 cpu coresmultilingual voice cloning48 khz stereo input and output support
key features
0.1b parameter tiny speech modelrealtime generation on 4 cpu coresmultilingual voice cloning48 khz stereo input and output support
spec & usage
supports 20 languages
ships with local web demo plus cli commands for generate and serve
onnx cpu path is nearly 2x faster than the original release in project tests
designed for cpu-first deployment without requiring a gpu
limitations
the repository launch notes said to treat licensing as pending until the root license file was published
scope:
audiovoiceresearchlocalgithubfreereal-time
launchApril 13, 2026
last releaseMay 6, 2026