From image to cutter: what actually works

In Cookie-AI only the cutter object type accepts an image reference. Stamps run on text because readable lines come more stably from a description than from a random photo. If you want to cut a logo or silhouette, stay on cutter and attach a clear reference.

The image mainly drives the outer outline. Soft shading, hair strands and background clutter do not become useful walls. The pipeline looks for a closed shape to cut — not a photoreal rebuild. Prepare the image as if you were cutting a stencil.

Credits apply here too. Trying a bad photo costs the same as a prepared one. Two minutes of cropping and boosting contrast often save a whole generation.

Black on white works best, or a dark shape on a light ground. The edge between subject and background must be sharp. Soft gradients, shadows under the object and similarly bright colours fray the contour. Then you get holes in the wall or runaway borders.

Think of the shadow outline again. If you can trace the subject with a finger without guessing where the rim is, contrast is good enough. If you hesitate at every step, the image needs more work: threshold, redraw, or a simple vector graphic.

Interior holes — eyes, letter counters — only when they are large and clear. Tiny holes become break points. For lettering in a cutter: prefer the outline of a shield over delicate glyphs, unless the letters are large and bold.

A phone shot of a fridge magnet against a tiled kitchen is almost always the wrong reference. Grout lines, reflections, other objects at the edge and soft focus compete with the subject. The pipeline cannot reliably tell what “the object” is.

Fabric, grass, wood grain and busy tablecloths interfere too. The background should be boring — ideally plain white or light grey. If the subject casts a shadow, light it flatter or remove the shadow in an editor before upload.

Group photos and collages are a poor fit. One subject per image. Several figures on one frame often fuse into one contour or many useless islands. Cut one figure out and upload only that.

Crop tight around the subject, but leave a little white margin. Extreme cutoffs can clip edges; too much surroundings bring chaos back. Centre the subject and keep the frame roughly square when you can — that matches typical cutter proportions better.

White or very light and simple is the best background. A transparent PNG with a clean alpha edge often works if the silhouette is unambiguous. Remove soft glows, drop shadows and filter noise.

If you only have a poor photo, redraw the silhouette: a black filled shape on white in a simple drawing app. The “uglier” image is often better for a cutter. Photorealism does not help here.

Clear line art or a filled silhouette beats almost every photo. Drawings already have the contrast and closed form the printer needs. Photos must be cleaned first — and still fail when the shape is organically soft or lit in strong 3D.

Angled side views warp the cut: a star shot from the side becomes a parallelogram. Shoot flat from above or use an orthographic drawing. For logos, use the original file or a screenshot without UI chrome and without shadows.

Colour photos are allowed, but colour is not an advantage. What matters is brightness difference. A red subject on a green ground can look similarly bright in greyscale and vanish. Check the reference in greyscale before you upload.

Do not upload protected characters, logos or marks you do not own. Disney heroes, game mascots, sports crests and third-party brand marks are off limits — even for “home use only” cutters. Cookie-AI is a tool for your own or clearly licensed artwork, not someone else’s franchise.

If you need a company logo for yourself, use only marks you have rights to. The same rule applies to gifts and clubs: no permission, no upload. Your own sketches, public-domain shapes and clearly licensed graphics are the safe path.

Legally and practically: a problematic subject can cause trouble and often makes poor geometry, because fan art is full of fine lines. Simple original silhouettes print better and stay on the right side of the terms of use.

Is the object switch on cutter? Is there exactly one subject? Is the background white or plain? Is the edge clear in greyscale? Are shadows, textures and UI borders gone? Is the crop tight? Do you own the reference — or is it clearly licensed?

If anything is missing, fix the image first. A black-and-white redraw saves more generations than ten credits on the same soft photo. The preview should show an open frame with a continuous wall — not a holey spike and not leftover background glued on.

In short: high-contrast subject, quiet ground, tight crop, your rights. Then the image becomes a cutter you can actually use — not just a file that looks “somehow similar” in preview.