Speak, or play a call into your mic — Binah transcribes and
checks every figure against the filing the moment it's said. The live
copilot: audio in, said-vs-filed out.
A ticker is required. Binah pulls the real filing first, then checks the live transcript against it.
Heard — live transcription
Binah — said vs. filed, live
Transcribed with a self-hostable engine (faster-whisper), then
reconciled by the same LiveCallSession engine. Needs a microphone and the voice
deps installed — the audio is processed in your engine, nothing about it leaves.