Talk Timer

Tracks how long each person in a conversation talks. No transcription — just minutes. Voices are clustered by pitch and timbre on the device. Audio is never sent anywhere.

idle total 00:00

mic

silent —

Settings

silence threshold 0.012 cluster looseness 1.4

Higher silence threshold = needs louder voice to count as speech. Higher cluster looseness = harder to create new speakers (good for similar voices).

Speakers

Tap a speaker tile while someone's talking to manually attribute speech to them.

Timeline (last 60s)

Works best when speakers have noticeably different voices, take turns (no crosstalk), and the room is quiet. With similar voices, raise cluster looseness; with crosstalk, take the results with a grain of salt.