Fluency Lab | Devpost

Inspiration

It is often challenging to judge yourself when preparing for an important presentation, interview, or pitch. Speaking from experience, practicing alone rarely provides objective or actionable feedback, especially on delivery, pacing, and confidence. We wanted to build a tool that helps people improve how they communicate by providing clear, data-driven insights into their speaking habits.

What it does

Fluency Lab helps users practice and improve their communication skills through three main modes: Interview, Presentation, and Elevator Pitch. Interview mode provides the user with 1–5 sample questions, records their responses, and delivers detailed feedback on clarity, pacing, structure, and confidence. Presentation mode allows users to upload a slideshow or script and receive feedback on delivery, engagement, and overall presentation effectiveness. Elevator Pitch mode challenges users to deliver a concise pitch under a strict time constraint and provides constructive feedback to maximize impact.

How we built it

Fluency Lab records user video and audio through an integrated webcam interface. Speech is transcribed and analyzed to extract communication metrics such as speaking rate, pauses, and filler word usage. These metrics, along with transcript context, are analyzed using Google Gemini to generate structured, actionable feedback focused on improving public speaking and general communication skills. User sessions and results are stored in MongoDB, and the application is deployed on cloud infrastructure.

Challenges we ran into

One challenge was reliably extracting meaningful fluency metrics from speech data in a short amount of time. Designing feedback that was both specific and encouraging required careful prompt engineering to avoid overly generic or overly critical responses. Balancing feature scope while delivering a polished end-to-end experience within the hackathon timeframe was also a challenge.

Accomplishments that we're proud of

We’re proud of building a complete, working system that provides immediate, actionable communication feedback from a simple recording. Fluency Lab turns raw speech data into clear insights that help users improve with each practice session. We also successfully delivered a strong MVP with multiple practice modes within the hackathon timeline.

What we learned

We learned how much value can be unlocked by combining speech analytics with large language models for skill development. We gained experience building reliable speech-processing pipelines, designing structured AI outputs, and creating user-focused feedback systems under time constraints.

What's next for Fluency Lab

Next, we plan to introduce real-time feedback during practice sessions, add long-term progress tracking, and support additional speaking scenarios such as team meetings and impromptu speaking. We also want to refine fluency scoring, add visual feedback timelines, and continue improving personalization and usability.