Hacker News new | ask | show | jobs
Show HN: Lingle – Voice agent to simulate zoom-based personal language lessons (lingle.ai)
4 points by andrewfhou 76 days ago
In my opinion the best way to learn a language (outside of moving to a different country) is to get a personal tutor and have consistent 1 on 1 lessons. I used Preply or iTalki for this back in the day but had issues with flexibility and pricing.

I've been trying to simulate the experiences that I had on those platforms with a voice agent.

You can try a demo for free here (or watch a video of me using it on the landing page):

https://lingle.ai/tryout/lingle-showcase

Right now the agent can plan a lesson with visuals and a conversation focus, guides the user through the lesson - correcting and explaining things on a whiteboard if necessary, and asks relatively engaging questions. There is also a long term user model that builds over time, that maintains memory of the user's produced vocabulary, grammar, and skills, meant to make each lesson informed by the past and build on the user's knowledge (obviously this is still relatively fragile in practice).

The voice agent itself has actually decent latency, especially given that turn detection was a bit of a challenge given that language learners often speak slowly and struggle when they're thinking, so I had to build with that in mind.

Anyways I wanted to get some feedback on the product. I think it's quite a different take on language learning tools than those that currently exist.

2 comments

Feedback - the landing page demo file which links to https://lingle.ai/demo.mp4 doesn't seem to play, instead it results in a 307 redirect to the sign-in page.

Also I've personally never tried it, but there was a fairly popular video based language chat tutor program called Call Annie which seems similar (talk with a virtual tutor who manages plan/vocab/etc). I don't think its active anymore but it might be a good reference point for you in terms of product fit.

Thanks for pointing out the issue with the video, should be fixed now. Interesting product you sent, although personally I'm not a huge fan of these uncanny valley-esque AI avatar tutors. Might just be me though.
Yeah, I think it came out a couple of years ago, which in the world of generative tech is basically the Stone Age. So it definitely has that sort of weird almost early Pixar/Wallace and Gromit look.
Oh god, I just went to the Call Annie landing page and first thing I see is the freaky AI slop avatar...her face is the last thing I'd want to stare at while I stumble through my Japanese study session every day. I'd rather talk to the Duolingo owl I think. The lingle approach of just having my face on the screen is actually kinda clever. Maybe it could recognize by my expression when I'm confused and help me out lol. I can also see how I actually look during the conversation which imo is a big part of communicating.
Of course everything looks bad with 20/20 hindsight. it's better than 90% of AI slop products out there right now.

This was back when agents were referred to as NLP generative chatbots lol, and I can guarantee you at a minimum the devs knew how to properly handle something as simple as say connection pooling, whereas most slop products these days are just an absolute shitshow behind the scenes.

the classic defense of every failed product... "trust me, the plumbing was solid." users don't care. they saw a creepy AI face and closed the tab. you're essentially arguing that the titanic had excellent rivets.
https://letterbook.ai

wtf even is this. Another wannabe AI SaaS product, your arguments sounds like any other YC founder who drank way too much VC cool aid.

I think you mean kool-aid, which refers to the well known American fruit-flavored drink mix brand. "Cool aid" on the other hand is completely nonsensical.
What do you use for the facial / sentiment detection? How does that affect the decision making of the AI?
Hume has an API for realtime expression detection - confusion states get injected into the LLM, kind of useful for better turn detection if the user is struggling but honestly doesn't work that well right now