Canvas-based image processing toolkit. Apply parametric pixel effects — toasts — to images and video via CLI or browser sandbox.
Zero production dependencies. Native Canvas and ImageData APIs only.
| Package | Description |
|---|---|
packages/core |
Toast library — effects, cutters, composers, color utils |
packages/cli |
CLI — run, batch, list, publish, login |
packages/sandbox |
Browser GUI — Svelte 5 + Vite, live preview |
packages/api |
REST API — Hono + SQLite |
pnpm install
pnpm toast list# Apply a toast to an image
pnpm toast run hue-scan -i photo.jpg -o out.png --param hue=180
# Process a folder
pnpm toast batch hue-scan -i ./photos/ -o ./results/ --param hue=90
# Render animation
pnpm toast run hue-cycle -i photo.jpg -o out.mp4| Slug | Output | Key params |
|---|---|---|
hue-noise |
PNG (transparent) | hue 0–359, noizeDeviation 0–1 |
hue-scan |
PNG | hue, noizeDeviation, background (#hex) |
hue-cycle |
MP4 / GIF | channels[], fps, format |
mosaic |
MP4 | tones, hues, minRegionSize |
# 1. Extract per-frame segments
pnpm toast mosaic frames -i input.mp4 -o batch-out/frames/ --tones 6 --hues 6
# 2. Download tile assets
pnpm toast mosaic collect-assets -o assets/downloaded --count 24
# 3. Render
pnpm toast mosaic render --segments batch-out/frames/ --assets assets/downloaded -o result.mp4cd packages/sandbox
pnpm devLive canvas preview — load an image, pick a toast, scrub the timeline.
Full docs: docs/cli.md
toast list
toast run <slug> -i <input> -o <output> [--param key=value ...]
toast batch <slug> -i <dir> -o <output-dir> [--param key=value ...]
toast publish -n <name> -p <preview.gif>
toast login <username> <password>
Source image → ImageData (Uint8ClampedArray, RGBA)
↓
Cutters — extract pixels by hue / saturation / value / channel
↓
Layers — wrap pixel data + blend mode + opacity + effects
↓
Reducer — merge layers pairwise via Composer
↓
Composition — orchestrates pipeline, owns canvas
Colors are normalized to 0–1 internally, 0–255 in storage.
- Node.js 20+
- pnpm
- ffmpeg (for video output)