Category
Technical Deep Dives
29 posts
- Technical Deep Dives··8 min read
Why Sharpening a Cutout Draws a Halo Around Fur
Upscale a cutout of a long-haired pet and a bright rim appears along the fur. The model did not do it — an alpha-blind sharpening pass did, by averaging in the black that sits behind every transparent pixel. Here is the mechanism, the fix, and exactly what the numbers behind it were measured on.
Technical Deep Dives··9 min readWhich Blurs Can Be Recovered, and Which Are Simply Gone
Camera shake, focus miss, subject motion and noise all look like 'blurry' and only some of them are reversible. Here is what a deconvolution model can actually put back, the resolution trade the tool makes to run in a browser tab, and how to tell in advance whether a photo is worth the attempt.
Technical Deep Dives··9 min readThe AI Mask Is Only the First Draft: Inside Our Cleanup Pipeline
The neural network is only the first stage of a clean cutout. See how NSS validates the raw mask, protects openings and thin structures, restores full-resolution edges, preserves hair, removes studio spill, and branches safely for glass.
Technical Deep Dives··8 min readHow NSS Verifies a Transparent Export Before You Download It
A checkerboard preview is not proof that a file contains usable transparency. Here is how NSS keeps straight alpha intact, verifies the encoded format, decodes the export again, samples soft-edge pixels, and falls back safely when PNG integrity fails.
Technical Deep Dives··9 min readVideo Background Removal Is Back: How the New Recurrent Engine Works
Video background removal is back, rebuilt around a recurrent matting network that carries state from frame to frame — so the result does not flicker the way per-frame masks do. Here is how the new engine works, why it runs entirely in your browser, and where its limits are.
Technical Deep Dives··9 min readThe WebGPU Shader Bug That Broke Best Quality — A Debugging Postmortem
Best Quality background removal worked in some browsers and silently failed in others with a cryptic WGSL compile error. Here is the trail we followed to the root cause — a bug deep in a machine-learning runtime — and the one-line-idea fix.
Technical Deep Dives··8 min readWhy AI Upscalers Leave Seams and Soft Edges — And How We Fixed Ours
An AI upscaler can sharpen a photo beautifully and still leave a faint grid across it, or a crisp subject with a strangely soft outline. Both are avoidable. Here is what actually causes them, and the two fixes we shipped.
Technical Deep Dives··9 min readfp32, fp16 and int8: What Model Precision Means for Your Cutouts
Fast or Best Quality? The real difference is how many bits each number in the model uses. Here's what fp32, fp16 and int8 do to your cutout's size, speed and edges.
Technical Deep Dives··10 min readSegmentation vs Matting: Why a "Good" Mask Can Still Look Cut Out
A perfect outline still looks pasted when hair and fur need partial transparency. Here is why segmentation and matting are different jobs, and why the harder one costs more.
Technical Deep Dives··9 min readWhy Cutouts Get a Coloured Fringe — and What Edge Decontamination Does
That thin green or white halo around a cut-out isn't a bad cut — it's colour spill baked into the edge pixels. Here's why it happens and what edge decontamination does about it.