Tutorial

How to simplify colored drawing for tattoo (B&W + AI)

Q: Does AI preserve correct proportions in simplification?

Yes — model trained to preserve geometry. Eye distance, nostril position, chin angle — all kept 1:1 with reference. If reference photo has distortion (fisheye lens, etc), output preserves that distortion too.

Q: Works with photos of real people?

Yes, but with caution. Decalker preserves facial features — if client doesn't want identifiable face (deceased family, etc), review first. For faithful portrait, high quality photo + Contour style = best result.

Q: Does reference color affect output?

Indirectly. Saturated color areas (intense red, black) tend to become well-marked stencil areas. Pastel/light areas become lighter lines or disappear. AI decides what's contour based on CONTRAST + geometry, not specific color.

Q: How many times can I regenerate?

Unlimited within free tier (10/mo total) or Pro (unlimited). Regenerating counts as 1 Decalker of month. For 1 photo, usually 2-3 regens give final version. Refine does NOT count as separate regen (it's edit of same piece).

Q: Tip for low-quality photos?

Decalker has CLAHE preprocessing that improves light/contrast automatically, but doesn't recover lost resolution. If client brings small photo (<600px): ask for new one with modern camera in natural light. For historical photos (old family), accept more simplified result and use Refinar to add details via text.

Time:5 seconds (Decalker) or 30-40 min (manual)Difficulty:Beginner

TL;DR

Colored photo has MUCH visual noise (shadows, textures, gradients) that confuses stencil. Decalker AI isolates ONLY what matters for tattoo — main contours + anchor points + structure. No shadow, no color, no noise. 5 seconds vs 30-40 min manual simplification in Photoshop.

Materials needed

•Colored theme photo (1500px+ ideal, good light)
•Decalker (free 10/mo enough)
•Image viewer for review (browser, Procreate)

Step by step

1. Identify what's EXTRA in colored photo

Typical colored photo has: main subject + background + light shadows + skin/paper texture + reflections + varying color. For tattoo, you only want SUBJECT + main internal structure. Mentally identify: 'shadows out, background out, texture noise out'.

2. Upload photo to Decalker

Decalker -> Decalkar -> drag photo. AI model (Flash 2) was trained to identify 'subject vs noise' in thousands of tattoo examples. Works better with focused photos (subject takes 60%+ of frame). Blurry/distant photo = bad stencil.

3. Choose right simplification style

Contour: simplifies MORE (only main contour, best for realism where you paint everything on skin). Anime: simplifies less (keeps internal structure for fillings). Sketch: selective simplification (keeps only 'essence' of drawing). Default Contour works for most cases where you want max reduction.

4. Check quality score + review output

Quality score >=75: ideal simplification — piece is clean, no residual noise. <60: reference photo had too much noise or was poorly lit — regenerate. 60-74: acceptable but consider regenerating. Manually check: main contour preserved? Essential details (eyes, nostril, contour) there? Shadow really gone?

5. Refine if more simplification needed

Decalker output still complex? Use Refinar with 'simplify center area' or 'remove background detail' or 'reduce line count by 30%'. Multiple iterations possible — each refine saves as new version in library for comparison.

Common mistakes

!Reference photo with flash (creates reflections AI picks as line)
!Photo with multiple subjects (AI picks random — focus 1)
!Expecting simplification keeping ALL color (doesn't happen — it's B&W stencil)
!Accepting first output without comparing to other generations
!Low quality photo (<800px) — model misses critical detail

Decalker shortcut

Manual Photoshop to simplify colored photo to stencil: 30-40 minutes. Decalker AI does it in 5 seconds with v2 pipeline (CLAHE preprocess + Flash 2 + Sauvola threshold + skin-physics-aware mode). 240x faster difference.

Try Decalker free

FAQs

Does AI preserve correct proportions in simplification?+

Yes — model trained to preserve geometry. Eye distance, nostril position, chin angle — all kept 1:1 with reference. If reference photo has distortion (fisheye lens, etc), output preserves that distortion too.

Works with photos of real people?+

Yes, but with caution. Decalker preserves facial features — if client doesn't want identifiable face (deceased family, etc), review first. For faithful portrait, high quality photo + Contour style = best result.

Does reference color affect output?+