The original pipeline began with yt-dlp: give it a channel or a playlist and it would fetch the episodes itself. That was convenient and it is how the first batches were made.
The decision
From 2026-10-04, Clipper works from files creators provide: a shared drive link, an upload, a file handed over as part of a campaign. The yt-dlp-based source types (free source URLs, creator feeds, discovery) are switched off by default, and a campaign source that happens to be hosted on YouTube is never downloaded.
The architecture diagram in the project’s documentation was redrawn to say so plainly: one owner, one computer, no server, and no data about any other YouTube user.
Why
Clipper’s relationship with YouTube is through the official Data API, used for uploading and for the owner’s own analytics. Keeping the input side on the same footing, footage the creators chose to hand over, makes the whole system simpler to describe and defend. It also matches how the campaigns actually work: the creators who want clips made are the ones supplying the episodes.
What it changed in the app
The Sources page now distinguishes campaign sources from the disabled fetch types, and the tick skips anything that would require a download. Existing clips made from downloaded episodes were not removed, but no new ones will be made that way.