search

uezo / ChatdollKit

ChatdollKit enables you to make your 3D model into a chatbot
Apache License 2.0
790 stars 79 forks source link

Voice recogintion #375

Open noriando opened 1 month ago

noriando commented 1 month ago

I had an issue with voice recognition, not only on iOS but also on WebGL. I spent many hours troubleshooting this, adding debug code, and making modifications.

To investigate, I downloaded the latest version (0.83) to clarify the issue, but ChatdollKit still doesn't recognize the voice.

Finally, I discovered that if I set the language field in the OpenAI Speech Listener to null and clear the cache, it starts working again.

If you can fix this issue, it would be a great help! I couldn't identify the cause from the OpenAI Speech Listener source code, but I suspect this is where the issue lies.

uezo commented 1 month ago

Hi @noriando , Thank you for investigation😊 What did you set to the language field? It should be ISO-639-1 format(e.g. en for English, not like en-US.)

https://platform.openai.com/docs/api-reference/audio/createTranscription

noriando commented 1 month ago

I set ja in language field.

now I realize it might be microphone conditions. In case I change headphone in stead of Mac microphone, it looks like ok. I continue to investigate.