sentis-whisper-tiny won't work to begin with

#9
by FlatHIppo - opened

So I was originally creating my own way of doing this but of course the API request is slow as heck so I found this and began trying to integrate it. Obviously its outdated so I needed to install the old version of sentis. After following all of the instructions, putting everything in place and the script on the camera and running it, heres the output:
image.png
I am using the audio file that is included, so that shouldnt be the issue. Just confused as to why it just says [Bell] instead of the transcription of the "voicemail". If theres anyone with knowledge on this and could help I would be forever grateful. I don't have a ton of experience with sentis so I don't know what I would modify to get this working.

I am using the exact script included. (https://paste.ofcode.org/3b5qp54jVdDd4wsQ6PQBhYt)
Yes I have moved the models (.sentis files) into StreamingAssets.
Yes I have moved the vocab into StreamingAssets.
Yes I have installed Newtonsoft.Json
Yes I have installed Unity.Sentis (com.unity.sentis) version 1.4.0-pre.3
Yes I put the script on the Camera.
Yes those print statements are called from line 181 where the output is printed in the RunWhisper file.

EDIT: I am using Unity 6. Maybe this is just not supported. I'll have to find another way to do this I fear.

Sign up or log in to comment