MeetBridge

Browser-based meeting translation copilot for PT ↔ EN.

Practical MVP for real calls: clear mode selection, a dedicated latest phrase area, short recent history, quick meeting phrases, and explicit provider status so you always know if you are seeing real translation or the built-in fallback.

Browser speech recognitionChecking provider statusNo external key required to run
Idle · microphone off

This browser is missing live recognition support.

Latest phrase

English speaker → Portuguese listener

Loading provider…en-US
Latest recognised phrase

Your latest final recognised phrase will appear here.

Latest translation

Translated output for the latest phrase will appear here.

Live preview while you speak

Interim transcript preview appears here while the browser is still listening.

Preview translation

Preview translation appears here for in-progress speech.

Loading provider details…

Running transcript

Meeting capture

Recent final phrases only

Final transcript chunks will accumulate here as the meeting progresses.

Recent history

Short bilingual turn log

No final phrase history yet.

Quick phrases

Meeting helpers

Ask for a repeat

Could you repeat that more slowly?

Pode repetir isso mais devagar?

Confirm deadline

Can we confirm the deadline?

Podemos confirmar o prazo?

Summarise actions

Let me summarise the action items.

Deixa-me resumir as ações.

Need clarification

I need a quick clarification before we continue.

Preciso de um esclarecimento rápido antes de continuarmos.

Translation path

How this MVP decides what you see

  • Browser speech recognition produces interim and final transcript chunks.
  • Final chunks are stored as recent meeting history; the latest phrase gets a dedicated translation area.
  • If LibreTranslate is configured, MeetBridge uses it as the real provider.
  • If not, or if it fails, MeetBridge falls back cleanly to the built-in local translator and shows that clearly in the UI.
Browser limitations

Important MVP constraints

  • This remains browser-based, so live transcription depends on the Web Speech API, not a dedicated STT backend.
  • Chrome and Edge desktop are usually the most reliable for this MVP.
  • Accuracy still depends on browser engine, microphone quality, accents, overlap, and background noise.
  • There is no speaker diarization, secure persistence, meeting import, or production-grade audio pipeline yet.