MultiverseEdTech
CarwowAuto
TicketmasterLive events
HablaApp
Adviser ProfilesBusiness
iOS App

Habla

Learn a language by speaking it. AI conversation partner for people who want practice, not gamification.

play it again
Year
2026
Role
Design & Development
Platform
iOS (SwiftUI)
Status
In development

A voice-led Spanish learning experience rebuilt as a native iOS app with one meaningful upgrade: you respond, and it listens. No tapping. No quizzes. Just listen, think, speak, repeat.

The problem

Audio courses teach well but offer no interaction — you answer in your head and never know if you’re right. Language apps offer plenty of interaction but weak methodology — tapping exercises and XP loops don’t teach you to speak. The gap between the two is where mistakes get reinforced and confidence stalls.

The approach

Instead of inventing new pedagogy, I started with a proven audio teaching method and designed a single, meaningful upgrade: the app listens when you speak and gives immediate feedback. The entire interface is governed by a six-state machine — tutor speaking, waiting, listening, processing, feedback, transition — with every visual element, animation, and timing value derived from that one state. The design is almost invisible: a waveform, a state label, and a text area. Nothing else.

The hardest design work was in the details that “simple” demands. The pause between the tutor finishing and the microphone activating had to feel like a natural breath, not a technical gap. The feedback needed variety to avoid becoming noise — pools of randomised phrases, never repeating consecutively, with enough specificity to be actionable. And there had to be an escape hatch for the learner who knows they don’t know: a quiet “play it again” button that appears after two seconds, offering a way out without shame or timeout anxiety.

Outcome

Still in development. The core interaction loop is working — a functional prototype with real speech recognition, fuzzy matching, and the full six-state machine running against a test lesson. Timing, feedback variation, and state transitions have been tuned through on-device testing.

Next focus is the content pipeline: generating structured lesson scripts with AI, producing audio with ElevenLabs using a three-voice system (English tutor, European Spanish, Latin American Spanish), and expanding from one test lesson to a twenty-lesson course.

All projects