Repository navigation
Replies: 1 comment
|
@n4ze3m Thanks for developing tiny-voca! A community model PR would be very welcome! Please also try to reuse the existing framework runtime, modules, and helpers as much as possible. They handle error-prone lifecycle and memory layout issues and provide centralized performance optimizations. Of course, for tiny models, native code could outperform GGML (e.g., #1095). Overall, choose whichever approach performs best. Please keep the phonemizer model local for now. Looking forward to the PR! |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi! I'm the author of tiny-voca, a small English text-to-speech model:
I would like to add it to audio.cpp as a community model, and I can port everything myself: the model, the waveform synth, and the phonemizer.
tiny-voca does not use eSpeak. It has its own phonemizer, dizione (https://github.com/n4ze3m/dizione, Apache-2.0, Python only for now), so I would write that in C++ too as part of the port.
Two questions before I start:
I will follow model spec v1 and the validation style from the other community PRs.
Thanks for the project!
All reactions