Whatsapp Speech to Text

[altbert]

Hi everyone!
I've made a Whatsapp speech to text transcription using Whisper, Flask and Whatsapp-Web.js

Can be self hosted or using the OpenAI api.

Here it's the code Whatsapp Speech To Text

It was a project i was thinking to make, but just recently had the time to make.

[cijjas]

hey! Question i am using some snippets of code you have on your repository. I will begin saying im new to this and maybe im doing dumb questions. The thing is, i already know how to convert mp3 files into text with openai's api but i dont understand your SpeechToTextTranscript function works, i find myself unable to understand what is the url that should be called with the request.post. i thought at first its only the whisper api url but its not. im not able to make it work. Can you help me understand what should i put instead of apiHost or the url overall. I would have solved this if i was able to convert base64 wav files into mp3 but im not able too. Thank you for your great work, will be attending to any response!

[altbert]

Hi! The apiHost variable is the url of the Flask API that handles the transcription using Whisper. The code for this API can be found in the repo under api/api.py.
The url should be edited to point to the server that you are hosting this API. In the case of this repo the variable is set to point to the right container in the docker-compose.yml.

[cijjas]

okay thank you!

[cijjas]

And thank you for your work

[hoehermann]

Just for fun, I just did the very same here with whatsmeow instead of Whatsapp-Web.js. I am glad we have so many options how to use Whisper!

[altbert]

That's really cool!