Canvas & Presets

The GPU-accelerated compositor turns a raw screenshot into a social-ready image. Pick a preset, tune three sliders, export.

Built-in canvas presets

Nineteen ship in the box, and every one of them is editable. The compositor stores padding, corner radius, scale, position, inset and the full shadow per preset, so picking a preset restores that destination's whole look.

Social

• x-post — 1600×900
• x-header — 1500×500
• instagram-post — 1080×1080
• instagram-story — 1080×1920
• instagram-portrait — 1080×1350
• linkedin-post — 1200×627
• facebook-post — 1200×630
• youtube-thumb — 1280×720
• pinterest-pin — 1000×1500
• og-image — 1200×630

Aspect ratios

• 16:9 — 1920×1080
• 3:2 — 1800×1200
• 4:3 — 1600×1200
• 1:1 — 1080×1080
• 4:5 — 1080×1350
• 9:16 — 1080×1920

Auto, Original, Custom

• Auto — fits the canvas to the source
• Original — keeps native size with minimal padding
• Custom — any size from 100 to 4096 px per side

Backgrounds

Four modes.

• None — transparent canvas, exports as PNG with alpha
• Solid colour — picker or hex
• Gradient — linear or radial, 2 to 5 colour stops, angle 0–360°
• Image — the 46-image catalog, or any file from disk

Shadow

One configurable drop shadow: on/off, colour, opacity, blur, distance, and direction on a 0–360° dial.

Screenshot layer

• Padding: 0–40% of the canvas
• Corner radius: 0–48 px
• Scale: 10–200%, with X/Y position and a centre reset

Content-aware cropping

Two tools beyond a plain crop.

Balance

Horizontal, vertical, or both. Scans the edge pixels and equalizes uneven borders, so a capture with 12 px of margin on the left and 31 px on the right comes out symmetric in one click.

Inset

Crops symmetrically from any combination of sides. Negative values extend the image instead, repeating the edge rows, so a capture cropped one pixel too tight gets breathing room without a recapture.

Export

Additional targets: clipboard, drag & drop to any window, open in default app.

Automation

Every preset, background and export setting is also available via MCP tools. Agents apply them exactly like humans do.