Soft product shots and phone photos often look fine on the screen until you crop or print. This workspace runs sharpening in the browser so you push detail without handing files to a server.
Load an image
Tap here, or drag a file into this panel. PNG, JPG, and WebP work well.
Result
People crank every slider to the maximum, then wonder why edges glow or noise turns into gravel.
Sharpening never invents real detail. It increases local contrast so transitions look crisper. When you overdo strength on a noisy file, you amplify sensor grain along with edges. When radius is too wide, halos appear around text and product outlines. The fix is smaller moves and frequent checks at the size you plan to ship, not the zoom level you happen to use while editing.
Large images stress mobile GPUs; resizing elsewhere first keeps the preview responsive. Transparent PNGs keep alpha, yet heavy sharpening still changes how fringes look next to removed backgrounds.
Nothing here replaces a print ICC workflow or raw development. If you need noise control before sharpening, run noise reduction first, then return here for a lighter pass. When a deliberate blur pass belongs in your stack, run blur as its own step so radius stays explicit.
Files stay in your tab. We do not upload or store your pictures.
| Mode | Best when | Watch for |
|---|---|---|
| Unsharp | General photos, screenshots, mild softness | Halos if radius or strength climb too high |
| Laplacian | You want a punchy local contrast lift | Can feel harsh on flat gradients |
| High-pass | Texture-heavy subjects where you want grit | Noisy skies amplify quickly |
| Edge | Line art, UI captures, diagrams | Ringing near thin strokes |
Use Output for a clean full view, Source when you need to remember what you started from, Split when you want a draggable seam across the same pixels.
A 2048 px packshot headed for a 1080 px carousel often looks acceptable at full resolution yet soft once scaled. Park the split line over lettering or fabric weave so you see whether you helped or hurt the downsized version.
Retail teams in Manchester shipped a spring lookbook in March 2024 using the same trick: they sharpened at the final 1200 px width, not the 6000 px master, because halos vanish until you match the delivery size. Your mileage varies with subject matter, yet the principle holds for nearly every JPEG leaving a marketing folder.
Social platforms re-encode uploads. A crisp edge in your PNG sometimes turns mushy after their compression. Keep strength modest so the platform encoder has less ringing to chew on.
Strength scales how strongly the filtered signal mixes back into the source. Radius feeds the blur stage inside unsharp and high-pass passes so the mask targets wider or tighter neighborhoods. Laplacian and Edge modes lean on convolution kernels instead of that blur step, so radius does less for them while strength still controls blend intensity.
We recommend starting with Unsharp near mid strength and radius 1 or 2, then switching to Edge only when the subject is mostly lines. Heavily compressed JPEGs need gentler strength than fresh exports from design tools.
Photographers who batch studio stills often separate global tone work from edge work on purpose. Tone rides in curves or contrast tools, edges ride in a dedicated sharpen pass at the end. Mixing both in one aggressive step is how skin pores turn crunchy while sweaters stay dull.
Canvas-based math here mirrors what textbooks label convolution: each output pixel looks at its neighbors, multiplies them by a small matrix, sums the result, then blends toward your original based on strength. Unsharp adds a twist by subtracting a blurred copy first so broad areas move less than sudden jumps.
Quick calibration. Toggle Source and Output after each tweak. If you need more than three round trips, duplicate the file and halve your strength instead of stacking extreme values.
Practical answers for browser-based sharpening.
No. Decoding, convolution, and PNG export happen with Canvas APIs inside your browser. Closing the tab clears the pixels from memory.
Start with Unsharp at moderate strength. Phone JPEGs already contain sharpening baked in by the camera pipeline, so aggressive Edge or Laplacian passes often show artifacts faster than on a raw file.
High-pass emphasizes fine variation. Smooth gradients like clear skies contain subtle noise that becomes visible when you amplify high frequencies. Lower strength or switch to Unsharp with a small radius.
It will not recover true detail from blur caused by movement or shallow depth of field. You might increase perceived edge contrast, yet smear remains. For optical blur you need a different capture or specialized deconvolution software.
Sharpen at the final print dimensions when possible, keep strength conservative, and soft-proof in your print toolchain. This page exports PNG; CMYK separation and paper profiles still belong in desktop publishing software.
Laplacian uses a fixed 3x3 kernel, so radius chiefly affects Unsharp and high-pass paths where blur width matters. Edge mode also ignores radius by design.
