dB ranges for your VAD?

mapluisch / OpenAI-Realtime-API-for-Unity

Implementation of OpenAI's Realtime API in Unity. Easily integrate low-latency, multi-modal conversations via push-to-talk or text.

MIT License

14 stars 3 forks source link

dB ranges for your VAD? #1

Closed yosun closed 1 month ago

yosun commented 1 month ago

Can't seem to get your VAD to work - have you looked into Whisper VAD?

https://github.com/Macoron/whisper.unity

yosun commented 1 month ago

https://github.com/Macoron/whisper.unity/blob/275406258aca21fe7753cf0724a65f06fd464eea/Packages/com.whisper.unity/Runtime/Utils/AudioUtils.cs#L69

mapluisch commented 1 month ago

Hey yosun, thanks for sharing! That looks great - will use it to replace my basic VAD approach.

mapluisch commented 1 month ago

Hey yosun, just wanted to let you know that I've integrated Whisper VAD (to the best of my knowledge 😅) - hope it works on your end. You might have to tinker around with the threshs in the Audio-Prefab.

yosun commented 1 month ago

VAD seems to cut off the response voice before it starts?

Also it seems that longer than 30s or 1 minute voice files do not get sent? (hangs in "recording stopped sending audio")