Name the voices from the call app
Name the unnamed voices of a recorded call from who the call app showed as speaking on screen, only where the screen evidence is enough, one name to one voice
Name the unnamed voices of a recorded call from who the call app showed as speaking on screen, only where the screen evidence is enough, one name to one voice
Give turns ([{ start, end, who }] in seconds, such as tool:who-is-in-this answered) and samples ([{ at, names }] or [{ at, elements }] in milliseconds from the start of the recording, elements being the call window's accessibility elements), and app (the call app's name) when known. Only voices whose label is still unnamed (Someone (...), Unsure (...)) are laid over. Returns assignments ([{ voice, name, share, overlap_ms }]) and unnamed ([{ voice, candidates }]). With apply, returns writes (each a name_speaker call's arguments, evidence included) for the caller to send; nothing is written here.
Nothing: your AI starts it with your words.
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to name the unnamed voices of a recorded call from who the call app showed as speaking on screen, only where the screen evidence is enough, one name to one voice”