Skip to content
github-actions[bot] edited this page Nov 21, 2026 · 4 revisions

The thumbs stage (Pipeline) makes tabloid-style covers: 16:9 for segments, compilations and clips, and a 9:16 one for each Short. The model picks the frame, the face and the words; Faces gives it each named person's best frames and the found face box, and the cast library gives it photos of people talked about but not on screen. The war room's cover editor changes any of it by hand.

Layout

Every position, size, color and fixed word of the tabloid covers is data, LAYOUT in clippa/thumbs/enquirer.py; a show changes any of it in layouts/<show>/enquirer.toml ([paths] layouts, beside its Shorts screenshots), e.g. [colors] green = "#0b3d91" or [cover.inset] size = [200, 180], and clippa thumbs <stream> --compose-only redraws with no model calls. An unknown key is an error. Leave the file out for the default covers.

[thumbnail] teasers = false (per show) drops the cover's three side boxes so the photo fills the left: one big moment, no other stories to plug. [preshow] teasers = false does the same for every pre-show and one-off cover only.

Photos

Tabloid covers can use a portrait of someone the show talks about who is never on screen. Make a folder every machine can reach (e.g. photos beside streams on the NAS), add one head-and-shoulders crop per person named <Person>.jpg (the whole picture is treated as the face, so crop out watermarks and logos), and set [thumbnail] photos to it (a relative folder is beside clipper.toml, like [paths]). The model may then use one for a cover's face or teaser box; clippa thumbs <stream> --items s04 --teaser-photo "Hasan Piker" sets one by hand. The cast library keeps several labelled photos per person in the same folder and fills itself.

Fonts: covers and Short captions use Impact and Bahnschrift, both standard on Windows (on Linux, ttf-mscorefonts-installer and fonts-roboto-unhinted).

Clone this wiki locally