AidanTilgner
commented
2 years ago OpenAI Whisper just came out and looks extremely promising. If I can reasonably run it locally I'd like to try, I think the speech to text would work a lot better than what I currently have, plus I'd have more features like translation to play with. It looks like I'll still need Silero for TTS though so we won't ditch that,
The speech-to-text of the speech server is a little wonky as of right now. This needs to be fixed because accuracy of speech is going to be required for development. There are a couple options to this, autocorrect and training. If I can train the model that would be ideal, but the autocorrect is another possible option.