Introduction to Assembly AI API
In this Bubble tutorial video, I'm going to show you step one of how you can use the Assembly AI API to extract different speakers and the text that they say from audio. We're going to be using the API to upload an audio file and then get a transcript back but know who said what in the transcript. But before I launch into that, did you know that we have got videos that you cannot find on YouTube exclusively available to our members at PlanetNoCode.com.
Recap on Assembly AI API
Right, this is going to pick up on some early videos where I was using the Assembly AI API. And so if you need a bit of a recap on each of the individual steps, you can go back and check out those videos. But I am going to be explaining what's going on here, which is that I'm in the Bubble API connector and I've added in an API called Assembly AI. I've added in my API key into the authorization field, private key and header and I'm making a post request to the Assembly AI API and this is the end point here.
Setting Up the API Call
It is an action so I can run it in a workflow and I'm sending it as JSON. And within the body, I've got one body parameter that I've made dynamic and that is I have to provide Assembly AI with a public open, accessible audio file or video file for them to be able to fetch and turn into a transcript. So I've uploaded an audio file to the Bubble app storage and here is the link directly to it there. The only step that really done differently from my earlier Assembly AI videos is that rather than this value into body, speaker_labels is true.
Initializing the API Call
And so if I initialize this call, and this will serve as a good recap for how the Assembly AI API works, I get back an ID and check out my other videos for how you can get this all automated running through a webhook but right now I'm just doing it in the API connector to demonstrate all of the steps. So I'm going to copy the ID because this is the unique identifier for the transcript.
Retrieving the Transcript
So once Assembly AI has finished processing the transcript either you provide them with a webhook, which I'll demonstrate in other videos, or you go and you look for the transcript using this ID and so for this I'm just going to look for the transcript. So I'm going to go down to my get process transcript ID and this is all covered in the Assembly AI documentation but I've laid it out here in the Bubble API connector. So I'm going to paste the ID into there and then initialize the call.
Analyzing the Transcript Results
And this is where I get back my transcript. So you can see here that my transcript starts with "Hello my name is Bubble, I'm speaker 1" and then someone else says "Hello my name is Emma, I'm speaker 2". But if I scroll down to utterances, I can see that it begins to group them and so I then have in utterance number 1, I have "Hello my name is Bubble, I'm speaker 1" and then Bubble only shows you one example but if I go to raw data ooh it's going to be a long way down, where is it? We then have utterance 2 "Hello my name is Emma, I'm speaker 2".
Conclusion and Next Steps
So part 1, I'm just showing how to get the response back that contains the data in JSON for identifying different speakers. Stay tuned for part 2, I'm going to show you how you can start to process this through the Bubble database and get the different parts of your conversation and display them.