Ghost-Terms
commented
1 year ago @dessant This would be very useful for my purposes, please consider adding it.
Open dann-merlin opened 1 year ago
Ghost-Terms
commented
1 year ago @dessant This would be very useful for my purposes, please consider adding it.
I implemented a custom HTTP API solver option.
I don't get why you don't want to have PR's, but here you go. There's an issue instead:
Is your feature request related to a problem? Please describe. My fork basically solves these issues: #386 #306 by providing a generic interface to connect to via HTTP.
Describe the solution you'd like My fork works for me. It would be nice if you could take a look, if I did it all correctly. I was a bit confused for example how the storage system works and basically left it out altogether.
If you want to merge my branch, I can create a PR.
My solution basically just sends the audio file via HTTP(S) to an endpoint that can be specified. An API Key can also be specified, which is then sent as a HTTP Header. The language is also sent via a HTTP Header. The transcription of the audio is then expected to be received as plain text.
Also in my testing HTTP did not work in chrom(ium), but it did in firefox. It did work for both when using HTTPS (I used a nginx reverse proxy for that).
I have not tested other browsers.
Describe alternatives you've considered This seems like a very generic interface to me which makes this extension very extensible for any kinds of transcription tools.
Additional context I wrote a python script which implements an HTTP API backend using OpenAI's whisper (running on CPU using whisper.cpp). It's really simple, so I invite you to take a look to see what I mean with "custom HTTP API". The script could basically be changed to use any transcription tool.
I very much appreciate the extension btw!