Want to finetune your language for voice input?
The model must be whisper-tiny, whisper-base or whisper-small finetuned on your language. We do not support large, large-v2 or large-v3 as it is too big to run on most phones.
For the model to work correctly in the apps, it should be finetuned with the ACFT method. This allows the model to run with greater efficiency. If you use a model in the app not finetuned with this, it will work with long dictations (30s) but shorter dictations (under 15s) will exhibit infinite repetition or a long delay at the end. If you already finetuned a model on your language but need help with this process, please make a new issue and we can help.
This document is still a WIP, more details will follow eventually. If you're interested in this or have done this before feel free to make an issue