We edited 10 travel vlogs in one night with Claude Code

This week we cut, captioned, and rendered ten full travel vlogs — plus a short teaser for each — in a single overnight session. The footage was from Ubud, Bali. The editor was Claude Code, driven from the terminal, wired to an ffmpeg render pipeline.

The reason it's worth writing down: editing ten videos by hand is a week of work. Timeline scrubbing, cutting, matching captions, exporting, re-exporting when the export is wrong. We ran the whole batch as one pipeline and let it render while we slept — twenty deliverables, one night.

Nothing here is a fancy demo. It's the same automation-first approach I use for client systems, pointed at video instead of inboxes: describe the job, let the machine do the mechanical part, stay on the hook for taste and quality.

How the system works, in three steps

1. 01
   Describe the edit, don't scrub a timeline.

   Each vlog gets a plain-language brief — the cold open, the chapter beats, where captions land, the pace. Claude Code reads it and assembles the edit decisions as code, not as manual clicks on a timeline.

2. 02
   ffmpeg does the cutting and captioning.

   The pipeline hands the plan to ffmpeg — trims, transitions, burned-in captions, color, the teaser cut-down. Because it's code, all ten videos run the exact same recipe, so the batch stays consistent instead of drifting video-to-video.

3. 03
   Render on the GPU, overnight, unattended.

   Final renders run on an NVIDIA RTX GPU using NVENC hardware encoding, which is the part that makes "ten videos in one night" real rather than aspirational. The queue runs while I'm asleep; the files are waiting in the morning.

The honest caveat: this is a first pass, not a finished-forever button. I still watch every render and fix what taste catches. But the mechanical bulk of it — the cutting, the captioning, the exporting ten times — is handled. What used to be a week is now a review session over coffee.

Why it matters if you make content: the wall was never ideas, it was editing time. When editing stops being the bottleneck, a founder or a small team can actually keep a channel fed — because the boring part runs itself.