A month ago, I received a series of voice messages from a friend that lasted 15 minutes
Fifteen
Minutes
By the time I listened, I could have
- Taken a shower
- Prepared lunch
- Call him and quickly speak about what bugs him
- Write the code for this hackathon (which, in fact, I did during the time he was busy recording the second series)
I talked to my gf about it that day. How much I would prefer if people just called. Or wrote a text message.
Fast forward, I read the announcement here. And well, you can imagine what instantly came to mind.
What if stop word-filled monologues like this one could be transcribed?
What problem does Chattergram solve?
Let's talk more serious. More often than not one finds themselves in a situation in which they cannot listen to spoken text.
- Because they're in a silent part of the train (is that a thing outside of germany?). Or in a crowded office/public space. You could read - but not listen to - a message
- Maybe they are in a hurry and just want to skim through the text very quick
- Due to a disease that causes them to be hearing impaired
The sender could still record their message. Just now the receiver can choose whether to listen or read.
Overview of My Submission
Chattergram is a full stack Typescript app that transcribes voice messages into readable text.
It aims to make voice messages more accessible by giving the receiver a choice. Read it. Or listen.
I have prepared a demo under https://chat.q-bit.me/. It uses the remaining credits from my trial phase - feel free to give it a shot.
Submission Category:
Accessibility Advocates
Link to Code on GitHub
tq-bit / chattergram
Transcribe speech messages into written text
About The Project
Chattergram is a full-stack Typescript chat application. With a particular extra:
It transcribes other people's (english) voice messages into written text for you
I've created Chattergram for the Dev.to & Deepgram Hackathon. It stands in no affiliation to the Android app 'Chattergram for Android'.
Features
- Above all: Voice message transcription from recorded audiofiles
- User authentication
- Persistent chats
- Swift deployment with docker & docker-compose
- Typed data structures & OpenAPI specification under
/api/docs
- Last but not least: Light & darkmode
Non-features
- Accurate user login stati
- User profiling
- Chat rooms
- Automated CI/CD
- Automated testing
Demo
You can try chattergram under https://chat.q-bit.me/.
The demo will remain up till the 31. of April or until my free Deepgram credit expires. It runs on a 2GB DO Droplet.
(…
Additional Resources / Info
Attached in the readme.md file, please have a look here: https://github.com/tq-bit/chattergram#getting-started
Top comments (5)
This is really fun.
I think this is an amazing idea. Your UI skills are off the hook. I usually prefer voice notes but sometimes receiving VNs longer than 5 minutes just makes me want to lose my shirt.
My feedback is that the transcribed results are displayed twice per instance. Also some instructions on how to use the product would come in handy.
Keep on keeping on.
Hi Muri,
thanks a lot for your feedback. Funny enough I priorized systematic code over UI on this one.
Agreed as well. My WA status says 'Before you send a VM, consider calling' for around 4 years now.
About the duplicated results: Did you happen to send a message to yourself? Or to a second account?
And about the instructions, do you think Vue tour might to the trick? I'd be curious to give it a shot. pulsardev.github.io/vue-tour/
I think I sent it to myself. No wonder it was duplicated.
I've just discovered Vue tour right now actually, it looks like it can do the trick.
I've got the duplicated chat part fixed, users can now properly send messages to themselves (they weren't getting saved before).
I decided against Vue tour though, it would've come down to a maximum of two steps to be described. Instead, I've added a placeholder for when a user logs in and didn't choose a chat partner.
Thanks again for your input