youtube-techcrushlisted

# YouTube Shorts from an existing archive Most "AI YouTube automation" builds videos from nothing: stock footage, a synthetic voice, a scraped script. That approach is cheap to start and structurally weak — the output is indistinguishable from thousands of identical channels, YouTube's inauthentic-content policy targets exactly that, and there is no moat. This skill does the opposite. It assumes the creator already has hours of original footage and treats that archive as the asset. The scarce resource is not video generation; it is **knowing which 45 seconds are worth cutting, and what to say in the first two.** Everything here follows from that. ## The core loop ``` transcripts → candidate moments → human picks → hook → render → human approves → upload ``` Automate the first two steps and the fifth. Never automate the fourth or the seventh. Hook writing needs judgment; uploading needs consent. ## Never upload without explicit consent Render the clip, show it to the user, wait. Publishing is irreversible, it is the creator's name on the channel, and a bad clip costs more than a slow one. This holds even when the user has approved previous uploads — approval is per-clip. If a workflow seems to call for automated publishing, build everything up to the upload step and hand over the file. ## Step 1 — Get transcripts (free, and better than you expect) For any video already on YouTube, YouTube's own auto-captions are the best source: free, instant, word-level timestamps,