A telegram bot to help you figure out if voice and audio messages are safe-to-listen
WebMB describes misophonia as "(...) a disorder in which certain sounds trigger emotional or physiological responses that some might perceive as unreasonable given the circumstance. Those who have misophonia might describe it as when a sound 'drives you crazy.' "
This bot aims to help people suffering from this condition judge whether or not a voice message or audio file is safe-to-listen. This all started as a joke between some friends, one of whom suffers from misophonia. Shout out to @jvmcarneiro for the original idea.
Misophonia Bot listens to any audio or voice message you send him and gives you his best estimate on whether or not that message is safe to listen to. Along with a verbal response with his recommendation, he gives you a score on a scale of 0 to 1 on where he thinks the message falls (with 0 being a clean message, and 1 being a very unsafe sound) so you can make your own judgement.
The brain of Misophonia Bot is a Recurrent Convolutional Neural Network trained on scrapped youtube videos and 1000+ manually annotated voice messages from telegram group chats. The model was inspired by this repo and the training code will be available soon.
Rest assured that no personal data is ever logged or stored by the bot. You can verify that by checking the source code yourself.
Misophonia Bot is currently hosted for free on Heroku, so some delays in response can be expected should the demand rise.
Record or forward a voice message to Misophonia Bot and you'll get your response. It's that simple! :)
Distributed under the MIT License. See LICENSE for more information.
Gabriel Lefundes Vieira - lefundes.gabriel@gmail.com