Forum Discussion

🚨 This forum is archived and read-only. To submit a forum post, please visit our new Developer Forum. 🚨
Ritik21's avatar
Ritik21
Honored Guest
2 years ago

Quest 2 Text to speech in unity latency

Hi,
I am facing a latency in text to speech oculus.
Does anyone know how can I reduce it or make it more like real time text to speech? 

1 Reply

Replies have been turned off for this discussion
  • baroquedub's avatar
    baroquedub
    MVP
    2 years ago

    EDIT. sorry just re-read your post and realise you're asking about the TTS part, which yes - does have latency. The way to minimise that, if you don't need realtime synthesis is to use the Disk Cache (https://developer.oculus.com/documentation/unity/voice-sdk-tts-overview/)

     

    My original message:
    If you're sending audio to a server you'll get latency no matter what. Not only does the voice recognition have to work its magic but you additionally have the network to contend with. Only way to speed things up is by doing your speech to text processing locally.
    I use this asset: Undertone - Offline Whisper AI Voice Recognition which I'd highly recommend. You can choose which of the Whisper model to download, and have found it lightweight and performant. Good community and support on the discord too. You could roll out your own for free, if you have the time and know-how.
    Here's a demo of that asset in action (not VR, but I guess it could be) https://youtu.be/4L5fOOpISeQ