Project Relate

Google LLC

About this app

Project Relate is a limited access app designed to help people with ​​non-standard speech be better understood in conversations.

You teach the app to understand your unique voice and speech patterns by recording a series of audio prompts.

Once you’ve completed your recordings, Project Relate will learn to understand the way you speak better. You can then use Listen & Repeat modes inside Project Relate to help be better understood by people around you & Project Relate keyboard in all Android Apps to transcribe your short-messages & long-form text.

Ratings and reviews

It has been over two weeks since I recorded the required 500 phrases. But the app still remains unusable and says learning in progress, it has not sent me any personalised speech model or any notification that the model for me is ready. Several contacts with their customer support team, but still no reply, except we will pass on your message but no time frame for resolution is promised. Still cannot use the app, it just says record more cards, even after recording 500 cards! No support !

Great feature. I was in a car wreck last year and damaged my jaw, cheek, and lost most of my front teeth. Since the accident it's been difficult to pronounce words the same as before. Participating in Relate a handful of times actually reduced my level of irritation. It's tough to have to repeat everything multiple times before anyone understands. Great idea, great deployment. I'm eager to see what possibilities it develops into.

I was pleased with it on my Android 8 phone, so updated to an Android 11 thinking it would be better. It is faster, but I have now lost the ability to choose a playback voice. Why is that? I would like to give a higher rating, but this new problem is a disappointment. The other problem with the app is understandable, as it can't cope well with much background noise.

What’s new

Project Relate is not accepting new users at the moment. Existing users are still able to access and use their speech recognition models if they already have one.