Back again. Gotta be honest it’s a very impressive app. Well polished. Easy to use, fun and actually effective at making me learn. I’ve proceeded to the paywall, without running into any major hiccups, bugs or errors.
I was especially impressed with the AI tutor lessons. It works extremely well, not clunky at all in terms of the wait times between when the AI is speaking and when you can respond. It’s very fluid and almost natural like. This is easily one of strongest selling points.
My number one suggestion is also with the AI tutor. Look into the “range” of acceptable correct responses for the AI Tutor lessons. For example, part of a lesson was about “Bebe”. When asked to say Bebe, it kept thinking I was saying “Baby” in English, not Bebe in Spanish. So maybe the recognition of a correct response is a bit too tight, I’ve never ran into an issue like that with Duolingo before. And since it couldn’t understand that I was answering properly, I just got stuck in a loop of it asking me to say “Bebe”, I respond, it doesn’t accept the answer and then it asks me an almost identical question. This happened maybe 10 times before it finally accepted one of my responses. Maybe loosen the acceptable pronunciation a tad, and instead also add on the same page, the “% pronunciation accuracy” you have on one of the other learning types.
The only bug I ran into was that sometimes when I launch the app, both via browser and the downloaded bookmark, there can be no sound. (The sound design is amazing so I wanted it)
If you want me to do more testing past the paywall let me know, I’d be happy to.
Keep up the great work!
Login to reply
Replies (1)
Thank you. That bug where it misunderstands you is just terrible, I’ve had a number of different versions where that kept coming back.
It’s bad because it reduces beginner confidence and trust in the user and it’s also just frustrating to be caught in a loop like that, especially when you’re saying the right thing. A similar one I have, to a lesser extent, is code switching between dialects/accents.
This voice stuff is surprisingly hard in this context.
Doing a deeper dive and different implementation on it right now in hopes it’s more stable.