Localizing Video — AI skill for Claude Code

localizing-video

**Make a foreign-language video say what it actually says — in English — with the dub, the subtitles, and the on-screen text all translated and in sync. Locally, for free.**


Why this exists

Ideas move across a language border through a chain of middlemen, and the chain is lossy by construction. A well-known figure says something; a second language's commentators paraphrase it; the paraphrase hardens into "the thing everyone now repeats." By the time it reaches you, you're not reacting to what was said — you're reacting to a summary of a summary, and you can't go check, because the primary source is in a language you don't read.

This tool removes the border. It takes the artifact whole — the spoken track in full, plus the slides and the on-screen text — and renders it in your language without dumbing it down. You get the primary source, not the telephone-game version. You can finally argue with the thing itself.

The two videos this was first built on make the point better than any pitch: both are Mandarin creators **dismantling** how Western AI buzzwords ("Harness", "Loop Engineering") get laundered — a famous name says it, influencers parrot it, it becomes a "new paradigm" nobody can define. That critique is sharp, and it was invisible to English speakers, who only ever got the hype. Localize it, and the deconstruction crosses the border too. A tool that breaks the language barrier, applied first to videos *about* breaking the hype barrier. That's the whole idea: **less friction between what someone said and who gets to understand it.**

What it produces

Three burned-in layers, in sync, from one foreign-language video:

1. Dubbed audio — neural TTS, each segment time-fitted to the original (with drift catch-up, since English runs longer); original-language clips left audible underneath wher