Interview to Clips - Extract Best Moments Automatically

Two speakers, one vertical frame

Interviews are the format where reframing earns its keep. 96% of our output is 9:16, and fitting two people into a vertical frame means either shrinking both until neither reads on a phone, or following whoever is speaking. We do the second, switching as the speaker changes.

On yield, a 15 to 30 minute interview returned an average of 5.4 clips across 462 measured sources as of 27 July 2026. The question-and-answer structure helps here in a way rambling monologue does not — a question marks a clean start point, and the answer usually resolves into something that stands alone without setup.

The failure mode to plan around is crosstalk. When two people talk over each other, speaker switching gets indecisive and transcription accuracy drops below its 0.972 average. Remote interviews on separate microphones come out noticeably cleaner than two people sharing one room mic, which is worth knowing before you record rather than after.