ufal / whisper_streaming

Whisper realtime streaming for long speech-to-text transcription and translation

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Enable translate to any language

prajdabre opened this issue · comments

Recently this hack was discovered to make whisper translate from X speech to Y text: https://github.com/Vaibhavs10/translate-with-whisper

Enabling this in whisper_streaming would be good.

good good !

I don't know if you're dealing with the form of the stream What is the data type of a variable, can you help me?

sorry, we can't. This is not related question to this issue and you don't give sufficient details

so, I read README of https://github.com/Vaibhavs10/translate-with-whisper , and it looks very simple: use audio in X language, and set Whisper as transcribe in Y language. This is already available in the current implementation of Whisper-Streaming. But I noticed that this zero-shot translation performance is rather small. It could be improved by tuning the parameters :

          "penalty_alpha": 0.6, 
          "top_k": 5,