pixels percent inches

Tselem found work from a previous session that wasn't saved or exported.

Fills the selected area with whatever you describe, using the surrounding image as context. The result arrives as a new layer, so your original stays untouched.

Runs a Real-ESRGAN super-resolution model in your browser — free, offline, nothing uploaded. The visible composite becomes a new document at 4× resolution. Tuned for illustrations, logos, and flat art (photos come out slightly stylized).

Grows the canvas and lets AI continue the scene into the new space. Opens as a new document — your original stays at full resolution on its own layer above the expansion.

Finds the subject automatically, then generates a new scene around it — the AI sees your product while painting the background, so lighting, perspective and shadows come out coherent. Delivered as two layers: the scene, and your subject cut out at full resolution on top.

Removes the selected object and rebuilds what's behind it — structures, edges and patterns continue through, not just texture. The result lands on a new layer covering the object.

Reads text that is baked into a flat image, lifts it out of the background, and hands it back as a real, editable text layer — matched font family, size and color. Select the words first (marquee or lasso), then Grab. Double-click the new layer afterward to retype it.

Traces the active layer into resolution-independent vector shapes. From then on, Image ▸ Image Size re-renders it razor-sharp at any size — vectorize a small logo, resize the document to 4500×5400, print crisp. Best for logos, text and flat art; photos become posterized poster art (for photos, use AI Upscale instead).

Morphs the active layer's actual letterforms — your font, size, warp and position — into a physical material. Lands on a new layer in place; the original text stays editable underneath.

Balloons Neon tube Gold Chrome Fur Flowers Fire Ice Cookie Wood Embroidery Stone & moss

Transforms what's already in the selection — recolor it, change its material, or repaint it in another style — while keeping its shape and composition. Lands on a new layer, soft selection edges respected.

The text

A vision model studies your composite and proposes a grade as real adjustment layers — Levels, Curves, Hue/Sat — each with its reasoning. Apply what you like, then double-click any of them to fine-tune. Nothing is baked.

A vision model looks at each layer's content and gives it a meaningful name — "Layer 7" becomes "Gold Pomegranate Illustration". One undo reverts all names.

Drafts marketplace copy straight from the artwork — title, description, accessibility alt-text, and tags. Copy each field into your listing.

Opacity 100% ⚡ Fast — Local
Texture synthesis in your browser. Free, seconds. Great for skies, grass, walls, patterns. ✨ Smart — AI
Rebuilds structure behind the object — railings, edges, geometry. ~4–6¢ per run.

Quality

—

Resolution metadata is stripped — web images don't need it.

Pattern description

One click, every size your workflow needs. Rows are remembered.

Tiled 3×3 — check the seams before saving:

Activity Lessons Load Server Log

Checked variants open as new document tabs, ready for Batch Export.

Returns real, editable text layers using your font library — including Hebrew faces. ~1¢ per run.

One tap applies — previews show your actual text. Everything stays live and editable (fine-tune later in Layer Style).

Plain resizing up softens detail \u2014 for a quality enlargement run AI \u25b8 Upscale first, then export at 100%.

Crisp vector edges for screen printing, engraving, and embroidery digitizing. Transparent areas stay empty.