Dennis H. Lewis walks through a Bubble app that records voice, sends the audio to Relevance AI for ChatGPT transcription, and returns the text in seconds. You can swap in different LLMs for the transcription step.
Chapters
We incorporated into this reusable element we recorded. We used a plugin for recording audio files. So there's a lot of them out there. And we'll put in our resource center a link to the one that we're using so that if you want to copy what we've built, then you can just do that.
Link to the resource center is in the show notes.
So we put that in as the recorder and then what we do is we take the URL from that file that, that the recorder has created and we send it to a backend workflow that we created with relevance AI that just uses, in this case ChatGPT to transcribe the audio file and return the text. As simple as that. You can use multiple different LLMs to transcribe text. That's up to you. You could swap them around in case, you know, if you're trying to generate long form content, well, maybe you want to use deep grammar. You know, there's, there's a lot of alternatives, but the process is seamless for the user. It's just talk into the microphone, hit stop, and keep going.
Let me summarize it for non technical users. So you've got this bubble plugin that essentially takes the user's voice input. We send that over to relevance AI. Relevance AI takes the audio file, it transcribes it using ChatGPT, it sends it back to our bubble.
Application, and we just put it into the text field.
How long does that take?
A couple seconds.
Couple seconds?
Yeah, it's pretty fast. I mean, it depends also on how long you talk.
