| This week we’ll have several people in the room with various tools attempting to “Vibe Code” our transcription service. We’ll pass screen share around to see each person’s approach and do some evaluation of the output. This should be interesting to compare with our existing transcription service that was written entirely by hand.
Here are some constraints to make sure it matches as much as possible to the tech stack:
- AWS for all hosting
- Docker Containers for everything
- Faster-Whisper for the speech to text model
- Flask back end
- React front end
- Stripe for payment processing
- Languages are Python and Typescript
- Bonus for CI/CD workflow in Github and Terraform deployment
Basic requirements
- User can upload a video or audio file, and set their email address
- User can pay, which starts the transcription process
- App updates the transcription live as it is being processed
- User receives an email with the transcription when complete
- App removes all containers used for processing as well as the file that was uploaded (a.k.a don’t keep any data)
- User does not have to create an account or log in
We’ll see how that goes in a live setting. If anything, it will be interesting to see how far some of these models have come. |
