Image emboss effect

You want a raised-metal read on a logo comp without opening a heavy editor. Load a raster here, swing the light around an eight-way rose, push depth until ridges feel honest, then grab a PNG. Nothing uploads; canvases read files the same way edge detection keeps gradients local.

Drop a frame into the vault

Tap anywhere on this tile or drag a file. Two canvases appear: one frozen source, one live relief pass you steer with the dock below.

PNG · JPEG · WebP plus anything else the browser decodes.

The first pass always looks like sheet metal

Emboss kernels subtract neighbors, add 128 to center the mid-gray, then clamp. Color survives unless you flip monochrome, which is why packaging comps often finish in zinc tones before CMYK separations.

We recommend previewing harsh logos on white and on charcoal slides. The same kernel reads bolder when surrounding luminance shifts.

Who moved the lamp?

Each rose button rotates a 3×3 stencil so highlights crawl along opposing edges. Diagonal lights exaggerate texture on fabric macros; top light keeps faces calmer on portraits you still treat as photos instead of badges.

When ridges feel mushy, nudge depth first. When halos look cooked, swing light before you blame the file. Aggressive sharpen stacked after emboss often rings; try emboss first, sharpen lightly, or skip sharpen entirely.

Depth slider
Scales kernel weights relative to the center tap. Low values whisper; high values carve. Extreme settings amplify JPEG blocks, so noisy phone shots benefit from noise reduction upstream.
Monochrome plate
Collapses RGB after the convolution so you see classic pressed-metal contrast without hue fighting the illusion.
PNG export
Grabs the relief canvas exactly as drawn, including alpha if your source carried transparency.
Privacy: Decoding and convolution run in your tab. We never receive the file. On shared machines, still treat sensitive scans carefully; browser memory persists until you close the page.

What emboss will not rescue

Vectors and tiny type blown up from social exports lack real high-frequency detail, so the filter invents slopes where none existed. Start from a native-resolution asset when the brief says “print ready.”

Animated GIFs decode to a single frame here. Video-style motion belongs elsewhere. For nostalgic warmth without ridge math, sepia shifts mood without pretending depth.

Very wide panoramas stress CPU loops on every slider move. Downscale a working copy, lock the look, then rerun on the master in desktop software if you need full resolution.

Stacks we have watched hold up

Brand teams emboss flat wordmarks, export PNG, drop into slide decks, then compare against foil mockups. Students print kernels for graphics homework without installing MATLAB. Makers preview leather stamp art before sending vectors to a laser shop.

Game UI artists sometimes emboss panel borders inside texture atlases, then downscale knowing the filter already exaggerated micro-contrast. Science communicators teaching convolution place a phone photo here, screenshot the pair, and annotate neighbor pixels in lecture slides because the visual pairs faster than equations alone.

After you like the relief, smart optimizer helps shrink PNG weight for email handoffs without reopening a separate compressor.

Numbers worth keeping straight

JPEG first, or PNG first?

Social JPEGs arrive with chroma subsampling and 8×8 quantization. Emboss treats block edges like miniature cliffs, so you sometimes see a faint grid on skies and skin. A PNG exported from the original edit, or a lightly compressed WebP from a clean master, keeps the illusion steadier.

When only a JPEG exists, drop depth one notch, try a gentler light angle, or run noise reduction before returning here. The goal is believable relief, not maximum slider drama.

Print designers still demand vector PDFs for foil dies; this page answers screen comps and student proofs where raster is the only artifact you have in hand.

Compared with outline tools

Edge detection answers “where do gradients change fast?” Emboss answers “how would this look if light raked across a thin plate?” The outputs look related only because both touch derivatives; the intents diverge. Use edges for masks, emboss for tactile fiction.

. Kernels follow the same directional family as classic emboss presets in reference texts; output matches the relief canvas above.

Relief, files, and odd edge cases

Short answers aimed at real previews, not textbook theory.

Why is the border pixel row slightly dull?

The 3×3 kernel needs neighbors on every side. The outermost ring never gets a full neighborhood, so values stay closer to neutral. Crop one pixel inside if the frame annoys you.

Does monochrome run before or after emboss?

After. Color participates in the convolution first, then the checkbox averages channels into gray so you keep luminance detail instead of pre-flattening color information.

Will this replace Photoshop emboss?

For quick browser proofs, yes. For layered documents, blend modes, or separations, desktop tools still win. Use this page when speed and privacy matter more than PSD history.

My file looks crunchy after emboss. Is the tool broken?

JPEG artifacts become slopes once neighbors differ across 8×8 blocks. Start from PNG, reduce depth slightly, or denoise first so the kernel is not amplifying compression noise.

Does Toolexe store uploads?

No upload occurs. FileReader pulls bytes into memory, canvases mutate pixels locally, and download builds a data URL. Network traffic matches loading any static page.

Why does top-left light feel different from bottom-right on the same portrait?

Each direction uses its own 3×3 weight pattern, not a simple mirror of one matrix. Cheeks, hair edges, and specular spots catch slope from different neighbor sums, so the emotional read shifts even when overall depth stays constant. Flip between TL and BR to compare before you commit.