Syllabal

• 

  Post-Conversation Analytics

• 

  App Dashboard

Inspiration

Kelvin's girlfriend uses Duolingo and while she's completing lessons, we noticed she wasn't improving in conversational fluency. Most language apps focus on vocabulary drills and grammar exercises, but they don't prepare you to speak. We wanted to build something that would help people go from "I know some words" to "I can hold a real conversation."

What it does

Syllabal is a conversation-first language learning app. You practice real-life scenarios (such as ordering food or booking a trip) by speaking with an AI companion that looks and feels like a real person. The AI listens, responds in real-time, and gives you immediate feedback on pronunciation, intonation, and pacing. You can chat with your companion in the language of your choice, and we track various metrics and provide suggestions to help you improve.

How we built it

We leveraged Digital Ocean's Gradient AI serverless inference for fast compute powering our conversational agents. The platform combines real-time speech recognition/transcription, natural language processing for contextual responses, and video-style AI avatars.

Challenges we ran into

Real-time transcription across multiple languages was our biggest bottleneck. Words that sound similar across languages, like "aéroport" in French being interpreted as a mispronounced "airport" in English, create ambiguity for the transcription system.

Accomplishments that we're proud of

We were able to build a complete end-to-end platform that enables real-time voice conversations with AI in multiple languages. Users can jump into a conversation in seconds, receive feedback on their speaking, and track their progress over time.

What we learned

Building the app reinforced the idea that learning a new language is hard work regardless of the tools available. But we once you remove friction (the boring drills, the anxiety of speaking to strangers), people are far more likely to stick with it and make progress.

What's next for Syllabal

We plan to fully flesh out the platform with more scenarios, languages, and personalized learning paths based on user goals. Ultimately, we aim to launch the app publicly to help people gain real conversational fluency in any language they choose.

Built With

• digitalocean
• python
• typescript