We’ve just added a new element that will enable more multimedia apps. Your users can now record sounds and save them. To do this, add the “Audio Recorder & Vizualiser” plugin to your app. You’ll then have a new input type called 'Audio recorder".
Once added to the app, you’ll have an element that can be controlled through 4 different actions:
start / stop recording
clear the recorder
upload (save)
play back.
You’ll have to make sure the sound is uploaded before using it in a workflow. The upload action returns a URL that will be what you use to save the sound to a thing. We’ve added some sound visualization stuff which makes the recorder quite fun to interact with
Kfawcett: I think that would require another extension in bubble if you dont want to “hack” it like you refer to.
If you feel the usability does not suffer from it then the current bubble implementation of the audio recording should be quite doable to connect to bubble.
That said, that flow would be:
Record audio file in browser with new bubble addon
Audio file transferred and saved to bubble backend
Sending audio file via API connector to watson or asking watson to download it from URL
Watson processing
Content from watson sent back into bubble backend
Bubble presentation of content back to end user
The downsides with this as per my opinion is that you could suffer a significant delay since the flow is this “complex”.
In the perfect world where you would want the transcribed content presented to the end user as fast as possible you would have a direct streaming of audio to watson which starts the moment you press record. Thus skipping probably at least 3 intermediary steps in the above list.
I am with you Gurun. I do not want to wait, I want the processing to happen in real time. I like the bridge you built, just in the spirit of Bubble, was wondering if it was something that could be “baked” in, instead having to use an HTML element.
So it would still happen from frontend directly to Watson, generating tokens on the backend and keeping that token in local state for you as a user when you send the data directly from your frontend to watson.
Maybe @emmanuel could give us an opinion of (approximately) how complex he sees such a bubble addon would be in terms of costs.
@emmanuel, see fifth post up. Examples for Google and Watson are provided that show how “Streaming” speech can be changed to text on the fly – go to either link, press the record/microphone button and start talking. Gurun was able to create a “bridge” to allow for this to work and pass the data to a Bubble input field, but it’s not something built in and requires an HTML element with javascript.