Why we built Auto Blemish with face detection, not a brush

The brush problem nobody talks about

Most photo editors assume you'll pick up a tool and learn how to use it. Brush mode, precision mode, strength slider. For professionals who move between Lightroom and Photoshop daily, that's second nature. For someone using their iPhone to shoot a selfie before uploading to Instagram, it's friction they don't need.

When we designed Cleanr, we made a bet early on: every feature should work in under three taps. Auto-enhance is one tap. Background removal is one tap. Sky replacement is one tap. So when we got to blemish removal, we kept asking ourselves the same question: why would we suddenly ask someone to learn brush control?

The answer was Vision face detection. Apple's framework can identify faces in an image and locate facial features with real precision. Instead of making the user do the locating and the cleaning, we could do both. Detect the face. Map the blemishes. Apply the fix. One tap, no learning curve.

What changed when we stopped assuming users wanted control

This sounds simple, but it meant rethinking what 'blemish removal' actually meant in Cleanr's design language.

A brush-based tool puts the user in charge. You see a spot, you paint over it, you adjust strength. You have control. You also have responsibility. If the result looks wrong, that's on you. You didn't hold the brush the right way.

With Vision face detection, we took that responsibility. We detect the blemish, we decide how much to blend, we handle the blending algorithm. The user taps 'Auto Blemish' and trusts us. That's a different contract.

It meant we had to get the algorithm right, not 'right enough.' We had to test on hundreds of face crops, different skin tones, different lighting conditions, different blemish types. A brush tool can fail and the user blames themselves. An automatic tool fails and the user blames us. The bar was higher.

But the payoff was obvious the first time we watched someone use it. No hesitation. No tutorial needed. No sense of 'am I using this correctly?' Just a clean result in a second.

Why this matters more than you'd think

Cleanr's primary users are faith creators and everyday people who shoot photos for Instagram, TikTok, Pinterest. They're not photography students. They're not enthusiasts. They're people who have one goal: make my photo look professional so I can share it and move on.

For that audience, a brush tool is a barrier. Not a feature. Not a creative option. A barrier.

When we shipped Auto Blemish, we watched what happened. Creator accounts started using it in their workflow. Not just once, but consistently. It became part of their daily routine, the same way they use Auto-Enhance or Smart Crop. They trusted it because it worked the first time, every time, without them having to learn anything new.

That's the whole philosophy of Cleanr, really. We use modern detection and processing frameworks so users don't have to. Portrait Blur uses Vision subject masking, not manual selection. Sky Replacement uses pre-calibrated presets, not manual colour grading. Auto Blemish uses face detection, not a brush.

The underlying technology is sophisticated. The user experience is simple. That's the goal.

What we learned about 'pro' features

Here's something we discovered while building this: calling something a 'pro' feature doesn't make it more valuable if it adds friction.

Some photo editors gatekeep blemish removal behind their highest paid tier, wrapped in language about 'precision control' and 'professional-grade tools.' The implication is that if you want real power, you need to pay more. What they're often really saying is, 'this is complicated, so we charge more.'

We made a different choice. Auto Blemish is available on the free tier. One tap, unlimited use. The paid tiers (Plus, Pro, AI Pro) add other things: batch processing, generative fill for advanced inpainting, face retouch for selective refinement. But the core ability to clean up a blemish, to make yourself look your best in a photo, that's free.

This decision came partly from listening to users, partly from watching how we use the app ourselves. Every one of us has a blemish we'd rather not see in a photo. Gatekeeping that behind a paywall felt wrong. Not evil, not predatory, just wrong.

The technical conversation nobody thinks about

Building Auto Blemish with Vision face detection meant solving a specific technical challenge: blemishes aren't uniform. A pimple is different from a scar, which is different from a birthmark. The algorithm had to distinguish between them and apply the right amount of blending.

We landed on a combination approach. Vision detects faces and facial landmarks. We then analyze the skin texture around detected blemishes to decide how aggressive to blend. Too aggressive and you lose natural skin texture. Too conservative and you don't clean up the blemish enough.

The result took weeks of testing. We ran the algorithm on selfies, studio portraits, outdoor shots, photos taken in low light. We watched the output and kept tweaking. There were moments we nearly gave up and went back to the brush approach, which would have been faster to ship and required less precision.

But every time we nearly quit, we'd remember that user message: 'I can't use blemish removal without learning a drawing app first.' That kept us honest. We shipped when Auto Blemish worked reliably. Not sooner.