RoughCut: An editing assistant for Turkish-language talking content.
It does the preparation and leaves the decisions to the editor.
01The problem
Cutting interviews, podcasts and talking-head videos means hours of listening for gaps, fillers and retakes, then checking that nothing meaningful got cut. When automation clips a syllable or breaks the context, the time it saved goes into fixing it.
02How it works
- 01
Analyze locally
Speech, silences and word timings are analyzed on your Mac. Your footage stays on your machine.
- 02
Propose, with reasons
RoughCut suggests cuts for silences, fillers and retakes. Each one shows why it was proposed and the exact source range.
- 03
Review in context
You hear every suggestion with what comes before and after it, then adjust the boundary or reject it.
- 04
Build a new timeline
Approved decisions become a new Resolve timeline. Your original edit is never touched.
03Beyond the cut
From the same transcript, RoughCut plans B-roll: where it belongs, why, and what kind of shot. It also suggests openings and selects for social clips. Everything it proposes is grounded in what was actually said, and nothing reaches the timeline without your approval.
04Why Turkish first
Turkish fillers like yani, şey and hani are sometimes noise and sometimes meaning. Existing tools are tuned for English. RoughCut is being built and tested on Turkish speech.
05Status
In development. Speech analysis runs locally; editorial suggestions are powered by Claude. I'm a working editor building this for my own projects first, then for a pilot with other Turkish editors.