metadata

license: apache-2.0
library_name: unity-sentis
pipeline_tag: automatic-speech-recognition

Whisper-Tiny model in Unity Sentis Format

This is the Whisper Tiny model tested to work in Unity 2023.

How to Use

Open a new scene in Unity 2023
Import package com.unity.sentis from the package manager.
Put the RunWhisper.cs on the Main Camera
Put the *.sentis files and the vocab.json in the Assets/StreamingAssets folder
Add a 16kHz mono audio file up to 30 seconds long to your project and drag on to the audioClip field.
IMPORTANT: The audio must be 16kHz. In the audio inspector select "Force Mono". And "Decompress on Load".
You can add a step to convert 44kHz or 22kHz audio to 16kHz with this model

When you press play the transcription of the audio will be displayed in the console window.