Transcripto - from talk to task instantly

Inspiration

We were inspired by the inefficiencies we all face during meetings—poor note-taking, missed action items, and forgotten decisions. We wanted to build something that truly listens and acts, transforming conversations into organized, actionable summaries. Our goal was to help teams focus on discussions rather than documentation.

What it does

Transcripto is an AI-powered assistant that allows users to upload or stream meeting recordings and automatically: a) Transcribes audio/video content b) Generates summaries in five formats (one-line, bullet-point, paragraph, in-depth, TL;DR) c) Extracts action items and decisions d) Analyzes tone and sentiment per speaker e) Displays speaker contributions and topic focus f) Lets users download all insights and outputs for future reference

How we built it

a) Frontend: Built with React for a clean, responsive dashboard, including filter dropdowns, scrollable panels, and download controls. b) Backend: Python-based APIs handling transcription parsing, action extraction, and tone analysis. c) AI Models: Used OpenAI GPT-4 for summaries and sentiment. d)Real-Time: Implemented live transcription using Streamlit and WebSocket for dynamic updates during meetings.

Challenges we ran into

a) Handling poor audio quality and speaker overlap in recordings b) Crafting effective prompts to consistently generate high-quality summaries c) Keeping the UI smooth and responsive with large amounts of transcript data d) Managing asynchronous tasks like uploading, transcription, and response generation without blocking the interface e) Operating within strict GPT-4 API rate and usage limits during the hackathon timeframe

Accomplishments that we're proud of

a) Successfully delivered a polished, end-to-end AI meeting assistant in under 48 hours b) Built a real-time transcription pipeline using Streamlit and WebSocket c) Developed multiple summary styles with GPT-4 d) Enabled users to interactively filter, explore, and download meeting insights e) Designed a sleek and intuitive frontend from scratch

What we learned

a) Effective UI/UX plays a crucial role in how users engage with AI-generated output b) Prompt design can significantly alter the quality and format of language model outputs c) Real-time systems require special attention to stream handling and frontend reactivity d) Good team coordination and clear task delegation were essential to shipping on time

What's next for Transcripto - from talk to task instantly

We aim to evolve Transcripto into a fully intelligent meeting companion by adding: a) Integration with calendars (Google, Outlook) for automatic meeting capture b) Native mobile apps for seamless on-the-go usage

Built With

• assemblyai
• django
• git
• github
• javascript
• openai
• pycharm
• python
• react
• streamlit
• vscode
• websockets