WhisperBoard

WhisperBoard — Turn conversations into live diagrams

Comment

Inspiration

We’ve all been in meetings where brilliant ideas were shared, but the key decisions and action items got lost in transcripts or people’s memories. WhisperBoard was created to solve this problem by making meetings visible—turning conversations into structured, collaborative diagrams in real time.

What It Does

WhisperBoard listens to conversations, transcribes them live, and automatically generates a collaborative board that organizes:

Topics → ideas being discussed

Decisions → choices that were made

Actions → who will do what, and by when

It also draws relationships between these elements, showing how ideas lead to decisions and decisions lead to next steps. The board refreshes continuously, becoming a living map of the meeting.

How We Built It

Attendee Bots → join Google Meet, Zoom, or Teams and stream transcripts via webhooks.

Backend Analysis → rolling transcript + lightweight NLP rules detect topics, decisions, owners, and deadlines.

Board Visualization → Miro API generates structured, color-coded sticky notes, connectors, and frames.

Frontend/Demo Layer → simple interface to launch bots, view live transcripts, and watch the evolving board.

Challenges

Noise vs. Signal → filtering casual chatter without losing key points.

Real-time Updates → debouncing & idempotent updates to avoid clutter.

API Coordination → synchronizing Attendee + Miro APIs under time pressure.

Layout Polish → ensuring diagrams look clear and professional.

Accomplishments

Built a system where a meeting “draws itself” live.

Pulled off real-time transcription → NLP → visualization in under 36 hours.

Designed outputs that impressed teammates with clarity and polish.

Balanced speed vs. polish to maximize wow factor in the demo.

What We Learned

How to integrate live transcription + webhooks in real time.

The power of visual storytelling vs plain transcripts.

That simple NLP + smart design can feel as magical as a full ML pipeline.

Hackathon insight: don’t over-engineer—prioritize demo impact.

What’s Next

Smarter analysis with LLMs for nuanced context and trade-offs.

Searchable memory → query past diagrams to recall decisions instantly.

Multi-modal input → unify audio, chat, and slides into boards.

Team integrations → auto-sync actions into Jira, Trello, Slack.

Long-term vision: WhisperBoard becomes the AI meeting memory where no decision or action is ever forgotten.

Built With

Share this project:

Updates