Speechi

Speechi is an AI-powered speech coach that listens to you practice and gives you instant, personalized feedback on your clarity, pacing, and confidence.

Comment

Inspiration Public speaking is one of the most universally feared skills — yet most people have no way to practice and get real feedback outside of expensive coaches or formal settings. We wanted to build a pocket-sized speech coach that gives instant, AI-powered feedback so anyone can improve on their own time.

What it does Speechi is a mobile app that listens to you speak and gives you detailed, personalized coaching. You record a speech or presentation, and Speechi analyzes your Clarity, Pace, and Confidence using AI. It also:

Transcribes your speech via AssemblyAI Suggests vocabulary improvements with specific word replacements Tracks your progress over time and adapts coaching to your history Offers structured practice modes: free practice, templates, and challenges Analyzes PowerPoint presentations to help you prepare How we built it React Native + Expo Router for cross-platform mobile AssemblyAI for accurate speech transcription OpenAI GPT-4o-mini for deep speech analysis and personalized coaching Clerk for authentication (Google OAuth) RevenueCat for subscription management AsyncStorage for local persistence and usage tracking NativeWind (Tailwind CSS) for styling Challenges we ran into Building a coaching context system that remembers user history and adapts feedback over time Balancing real-time feedback latency with analysis depth Designing a paywall and free tier that felt fair without blocking core value Getting audio recording to work reliably across iOS and Android Accomplishments that we're proud of A fully personalized AI coach that gets smarter with every session A clean, intuitive UI that makes speech practice feel approachable The vocabulary boost feature that gives concrete word-level suggestions, not just generic advice End-to-end flow from recording to actionable feedback in seconds What we learned Prompt engineering matters enormously — getting the AI to give specific, actionable feedback (not vague encouragement) took many iterations Mobile audio pipelines are surprisingly complex Users want progress tracking as much as they want feedback — seeing improvement is motivating What's next for Speechi Real-time filler word detection ("um", "uh") during recording Interview and pitch-specific coaching modes Social features — share recordings and get peer feedback Integration with slide decks for synchronized presentation coaching Voice tone and emotion analysis

Built With

Share this project:

Updates