stock-images · diff

git:20261011.4f4fc53 to git:20261012.a4c780a

3 added, 2 removed. Audit A to A.

---
name: stock-images
description: Find, download and place stock photos and stock video (Pexels) and vector icons (Iconify, as transparent PNG stickers) in a BashCut edit as illustration that blends into the cut, or use a blurred copy of the user's own footage as a background. Use when the edit needs a picture or icon the footage lacks, or the user says "ảnh minh hoạ", "video minh hoạ", "ảnh stock", "pexels", "lấy ảnh trên mạng", "thêm hình mô tả", "icon", "biểu tượng", "sticker minh hoạ".
---
# Stock photos and video
Reply in the user's language.
## 0. Library first
`bashcut library list --kind sticker --query "pot"` may already hold the picture. A plugin may search stock for
you: `bashcut library search "clay pot" --kind sticker` (a job). Place a library picture with `bashcut library place ID --at-frame F --size 0.6 --base-rev N`.
## 1. Search
- **Pexels API** (free key at pexels.com/api, 200 requests/hour): with `PEXELS_API_KEY` set,
`curl -H "Authorization: $PEXELS_API_KEY" "https://api.pexels.com/v1/search?query=<q>&orientation=portrait&per_page=15"`
(videos: `https://api.pexels.com/videos/search?...`).
- **No key** is not a reason to skip: open `https://www.pexels.com/search/<q>/?orientation=portrait` (videos:
`/search/videos/<q>/`) in a browser tool, read each result's link and alt text, and look at the thumbnails
before choosing. Vietnamese queries work.
- Photo ID = the trailing number of `/photo/<slug>-<id>/`. Download URL:
`https://images.pexels.com/photos/<id>/pexels-photo-<id>.jpeg?auto=compress&w=1080`.
## 2. Download
Download what the edit needs without asking, whatever the licence (photos ~100–250 KB at 1080 px; vertical stock
video 15–60 MB) into the project's `media/stock/`; record nothing (`--source URL` only when at hand), the user
handles rights. Avoid identifiable people. A picture worth reusing goes to the library:
```sh
bashcut library add --kind sticker --name "Clay pot" --file /abs/media/stock/x.jpg --tags food
```
## 3. Place
- - **No credits** on the video or in its description unless the user asks; then import with `--origin stock
- --license … --source URL --author "Name"` (a second `media import` adds them to a file already in) and write the
+ - **No credits** on the video or in its description unless the user asks; then import with `--origin stock` (or
+ `--origin web` for a file taken from someone's own page, such as a CC BY clip) `--license … --source URL --author
+ "Name"` (a second `media import` adds them to a file already in) and write the
lines from `project credits`.
- **Stock video** imports like any clip: `media import /abs/media/stock/x.mp4 --base-rev N`, then `media place`
on the main layer (it replaces a moment) or an overlay layer (it illustrates one).
- **Still photos** import directly: `media import /abs/media/stock/photo.jpg --place --track OVERLAY --at-frame F
--base-rev N` (kind `image`; JPEG, PNG with transparency, HEIC). An image is placed for 3 s; trim it to the words
it illustrates. Frame it with `transform` (zoom below 1 for a card, pan/tilt to place it), and give it life with
`clip motion ITEM --preset zoom-in` (or `pan-left`/`pan-right`): a still that does not move looks frozen.
Never convert photos to video files.
- **Generated B-roll** (only when a provider serves it: `capabilities get library.generate --kind clip`; none is
not a reason to skip stock): `library generate "<shot, framing, light>" --kind clip --request-id broll-1
[--params '{...}']` (paid providers: `--dry-run` first and say the price; model options such as length and aspect
are the provider's, read them from its options). Save the one the user picks with `library add --from-result
JOB:N`, then `library place ID --track OVERLAY --at-frame F --duration N --base-rev N`. It is AI picture: the
rules below apply as for stock, and `project credits` counts it in the AI share.
## 4. Rules
- **No labels on the picture.** Never write "minh hoạ", "stock", "AI", a source or a licence on the video: it makes
the cut feel unnatural. Keep the real footage as the main picture and let stock read as illustration by context
(on top of or between real shots, tied to the words), not by a caption. Never present stock as the real place in
the narration or titles.
- Tie each picture to the words it illustrates ("nồi đất" → clay pot); 1.4–2 s each, a soft pop sound.
- How many: the `stock` and `icon` rows of `plan.visuals` (`bc:visual-plan` balances them over the video). Without a
visual plan, 2–3 illustrations per video at most. Always off faces.
## Icons (Iconify)
For a small illustration (money, location pin, clock, check mark, arrow), an icon reads better than a photo. Iconify serves 200k+ vector icons through a free API (no key, ~0.2 s per call, tested
October 2026):
```sh
curl -s "https://api.iconify.design/search?query=cash&limit=20" # → icons ["mdi:cash", "solar:wallet-bold", …]
curl -s "https://api.iconify.design/collections?prefixes=mdi,solar" # → each set's licence
curl -s -o cash.svg "https://api.iconify.design/mdi/cash.svg?color=%23FFD400&height=512"
rsvg-convert -w 512 -h 512 cash.svg -o cash.png # transparent PNG; or: magick -background none -density 1536 cash.svg -resize 512x512 cash.png
```
- BashCut imports PNG, not SVG: convert first. Never use `qlmanage` for this: it paints a white background.
- Prefer filled sets (names with `solid`, `fill`, `bold`, or `fluent-emoji-flat` for colour emoji): thin outline
icons disappear on a phone. `?color=` recolours one-colour icons; colour emoji keep their own colours.
- Download without asking, like photos. Save under `media/stock/icons/`, place like a still
(`media import … --place --track OVERLAY`), 512 px, small (`transform` zoom ~0.25–0.4), with a pop sound, and
keep the good ones: `bashcut library add --kind sticker --name "Cash" --file /abs/media/stock/icons/cash.png --tags money`.
## Free fallback: blurred own footage
For a background (a horizontal clip in a vertical frame, a split moment, an info card), put a copy of the same
clip on the layer below, zoomed until it fills the frame, and darken and soften it with an adjustment until the
subject in front reads clearly. There is no right value: footage, text and grade all change it. As an example
only, zoom ~1.8 with `exposure` −1 and `saturation` 0.7 has worked; a bright, busy background needs more, a dark
one less. Check with `ui frame F` and `ui frame F --phone` (text and edges at the size a viewer sees them) and
adjust. Same colours, no licence question.