Inspiration
It started during a work trip to Cairo. Every day, my local colleagues would teach me a few words or phrases (usually things I had failed to say the day before, or ones I might need for tomorrow). I really enjoyed using these words to communicate with locals, I even said “Good morning” in Arabic about 800 times. Even if I messed up (and that happened quite often), I could just try again and learn it instantly.
But not everyone can do that. Most people freeze or hesitate, and when they travel abroad, they find it hard to fit in. That made me realize: people need a safe space to rehearse and build confidence before diving into a foreign language environment.
Then I discovered something weirdly perfect — Google Street View. Sometimes when you drop into a location, you don’t know where you’ll land. Once, I ended up inside a restaurant, sitting across from a man holding a drink and looking right at me. It was accidentally the most immersive language learning moment ever.
That randomness, that sense of “I shouldn’t be here but I am”, inspired Rehearsal.
What it does
Rehearsal turns Google Maps into an interactive language playground. Drop Pegman anywhere and start talking to AI-generated locals — each with their own accent, language ability, and personality.
Yes, I know you might think: “Oh no, not another language learning tool.”
But look, this one couldn’t exist without Google Maps. Street View gives a sense of place that no other platform can, because you actually feel the environment you’re trying to speak in. That’s where learning happens.
Rehearsal isn’t about learning words, it's about learning the world. It’s about recreating the real-world confusion and magic of landing somewhere new, encouraging you to explore, make mistakes, and keep talking.
How we built it
Honestly, I had no idea what I was doing. This project was built through vibe coding, curiosity, and some desperate conversations with AI.
I actually made two demos at once:
- One inside Google AI Studio (which worked well but wasn’t easy to turn into an extension)
- And one as a Chrome extension, using Groq’s free API instead of Gemini (it runs faster, but the dialogue quality isn’t as great, and I was using this was because I found it hard to figure out which Gemini API works).
The AI Studio version sometimes tried to skip the map entirely (though it did send me hundreds lines of apologies), while the extension respected the map view all the time, which somehow made the process easier. But it worked anyway, so the one you saw was in the form of web app.
Technically, Rehearsal combines, as my AI told me:
- Google Maps API – for world data and Street View rendering
- Gemini API – for context-based dialogue and image generation
- Speech-to-text / text-to-speech – for natural voice interactions
Challenges we ran into
- Location accuracy — I thought it’d be easy to let the AI “see” my screen and read the place name in the corner, but apparently AI doesn’t work that way.
- API madness — There are too many APIs, too many limits, and I still don’t know which one actually broke.
- Voice timing & realism — Teaching AI to react as a human being with a foreigner is not that easy, maybe because it basically knows everything.
- One-person dev life — I found this hackathon a week ago. That’s… not great for sleep.
Accomplishments that we're proud of
I built a working prototype that actually talks back. It’s buggy, unpredictable, and occasionally existential. But hey, it's alive. And somehow, it captures the exact feeling of being lost in a new country but trying anyway (and I laughed a lot while testing, which wasn’t planned but felt right.).
What we learned
Everything. Literally everything. From APIs and rate limits to Markdown syntax, including debugging panic attacks, every step was a crash course.
But most importantly, I learned that you don’t have to know everything to start building something that feels right.
What's next for Rehearsal — Practice for the real world
As partly mentioned before, I want to push immersion further including adding ambient sounds, enabling more real-data-based interactions (e.g., showing actual shop photos instead of AI-generated ones), or even creating mini-scripts with local characters.
On the marketing side, I’d love to see collaborations: celebrity cameos, pop-up stores, or virtual events hidden in real-world map spots.
There’s a huge space between learning a language and living in one.
That’s the space Rehearsal wants to play in.
Log in or sign up for Devpost to join the conversation.