Plan a video's stock B-roll — estimate

any = video first, a photo where no clip fits

cutaways: auto = portrait shots full frame, landscape ones as a split screen

Copy each chosen file to our storage (24 h) — a render never depends on another host, and Pixabay's terms require it

cutaways: how many shots

A light dark band behind the words over the footage (cutaways: the shot's half; faceless: the word band) so small words read on busy pictures

cutaways = 2–4 stock shots over a talking head · faceless = stock footage and photos over the whole video (the voice goes on)

One line on what the video is about

Show the best candidates' thumbnails to Gemini Flash-Lite before choosing (≈ $0.001–0.005)

ad = an advert or a sales video: no stock person (and no brand) is shown, so nobody in the footage seems to endorse the product — with verify, every chosen shot's thumbnail was checked for a face or a logo

cutaways: how long each shot stays

The output's length in seconds (the kinetic plan's summary.output_seconds); default: the last word + 1.5 s

Default: the edit's output, else 1080

split: where the stock shot goes (bottom keeps the face on top, the words over the shot)

Only where the server's licence rule allows more than CC0 (setting stock_auto_license = cc0-and-free): not-required = CC0 / public domain / Pexels / Pixabay; any = CC BY too, with its credit line on screen while it shows (ShareAlike never). Under the default rule every shot is CC0.

Default: the edit's output, else 9:16

faceless: a new shot about this often

split: which band of the speaker stays (0 top … 1 bottom; a selfie face ≈ 0.35)

Or a transcript whose clock is the output's (e.g. of the cut audio) — read here, its words in ms