Real time transcription

[easygoer33]

I have been using openwhispr for a while and love the app. However, I would like to know whether real time transcription is supported by openwhispr? I have so far assumed that when I turn on dictation or meeting mode, the transcript comes at the end of the recording after openwhispr has processed the audio. Is this correct? Is there a real time transcription feature, which would be huge?

[bernardtrier1975]

Great Idea, @easygoer33 !

I could imagine something like this:

A separate hotkey opens a foreground window and immediately starts recording. You can dictate, and the text is transcribed in real time so you can see it directly in the open window. This would also make it possible to manually edit the text right away if certain words were recognized incorrectly or if you want to rephrase a sentence.

Only when you click “Continue” or press the same hotkey again would the dictation enhancement / rewriting step start. The revised text could then be shown directly below or to the right of the original transcription. Even better would be a view where the changes are highlighted directly. With one click, the improved version could then be accepted and inserted into the original application.

I think this would be extremely useful, especially for longer texts. Real-time transcription would also likely reduce the perceived processing time significantly.

@gabrielste1n Could this be an idea for OpenWhispr?

[gabrielste1n]

this is pretty much possible for local models you will see underneath a toggle for "live dictation preview" if you toggle that on you will see a preview of your transcription as you speak.

[naborajs]

Here I was facing a problem that whenever I speak anything in another language other than English, like I speak Hindi or Bengali, then the thing happens is it understands that thing but it automatically writes transcripts it to English . Why does it happen? I say it in Hindi, then it should also write it in Hindi, right? So why does this thing happen? Can anybody have any idea to fix it or something?

[gabrielste1n]

can you please tell me what model you are using in STT settings?

[naborajs]

So currently I am using this OpenAi Turbo model. I saw it supports other languages but why is it translating other languages to English and then giving the output @gabrielste1n ?

[Clinteastman]

Open AI just released 2 new models:

https://developers.openai.com/api/docs/models/gpt-live-transcribe
https://developers.openai.com/api/docs/models/gpt-transcribe

Part of their demo is showing off it doing live translate of multilingual sessions.

https://www.youtube.com/watch?v=WeP9VUf1OoE