Why Browser AI Is a Compromise That Could Cost You Your Visa

We used to run everything in your browser. Privacy-first, no uploads, all processing on your device. It sounded right. It felt right. And for basic cropping and background removal, it worked.

But consulates don't grade on effort. They run your photo through automated checks. Highlight on your forehead above a certain threshold? Rejected. Uneven lighting across your face? Rejected. Subtle artifacts that your phone screen doesn't show? Rejected.

We shifted to server-side AI not because we wanted to — but because the browser could not deliver what immigration systems demand.

The Hard Limits of Client-Side Processing

Browser-based AI runs on WebGL, WebAssembly, and whatever neural network we can squeeze into a few hundred megabytes. It's impressive. It's also constrained.

Two problems we could never solve in the browser:

1. Compliance checks, not face editing

Overexposed forehead, nose bridge, or cheeks are among the top rejection reasons. Immigration systems (e.g. US) scan for pixel values above a threshold — typically around 235–240 in RGB. If your face has hot spots, the check fails.

We don't alter your face to fix highlights or shadows. U.S. and EU rules require a true likeness. Instead, we run the same photometric checks that consulates use. If your photo fails (highlights, uneven lighting, shadows), our Compliance Report tells you exactly what failed so you can retake with better lighting. Server-side we have the compute to run these checks accurately; in the browser we couldn't run them reliably.

2. What we actually do on the server

One side of your face lit by a window, the other in shadow — consulate checks often reject that. We don't "fix" your face; we crop, remove the background, and set dimensions to the published spec. We then run compliance checks (geometry, photometry, biometrics). If something fails, we report it. You get a compliant photo only when the source photo already meets the requirements — or you retake.

Why Server-Side Processing Matters

On the server we run background removal, crop and resize to the exact document spec, and a full compliance pipeline (geometry, lighting checks, biometric checks). We do not alter your face — no highlight "suppression", no skin "normalization". That keeps your photo a true likeness, as required by U.S. and EU document rules.

What server-side gives you that the browser couldn't:

• Accurate compliance checks — same logic consulates use (head size, background, photometric thresholds)
• Reliable background removal and crop to the published spec
• Compliance Report — if your photo has highlights or shadows that would cause rejection, we tell you so you can retake
• No face editing — we only crop, resize, and set background; your face stays as in the original

The goal is one thing: a photo that passes the consulate's automated checks — without changing your face.

That combination — strict checks + no face alteration — wasn't achievable in the browser. We moved it to the server.

Security: What Happens to Your Photo

We understand the trade-off. Processing on our servers means your photo is transmitted. So we built the pipeline around a simple rule: no retention.

• Transmission: All uploads use HTTPS (TLS 1.2+). Your photo cannot be intercepted in transit.
• Processing: The image is held in memory only. We run background removal, crop to spec, and compliance checks. We do not alter your face.
• Deletion: As soon as the response is sent, the original and all intermediates are discarded. We do not store, archive, or log your images.

This is not a marketing claim. It's an architectural constraint. Our processing pipeline has no persistent storage for user photos. By design.

What We Sell: No Problems at the Border

We're not selling "how it works." We're selling the absence of problems when you hand your documents to the consulate.

You care that your visa application isn't sent back because of a highlight on your forehead or a shadow under your chin. We run the same checks the consulate runs and tell you in the Compliance Report if your photo would fail — so you can retake with better lighting before you submit. We don't alter your face; we ensure compliance.

The architecture shift — browser to server — was made for one reason: to maximize the chance that your photo is accepted. Everything else — privacy controls, deletion policies, encryption — exists to make that shift acceptable.

If we could achieve the same result in the browser, we would. We couldn't. So we didn't.