A photograph of your dog, your wedding, your grandmother's garden — uploaded as an ordinary phone snapshot and returned days later as a numbered canvas with acrylic paints matched to every region. Custom paint-by-numbers kits are among the most tangible consumer uses of AI-assisted image processing. Festede, a Budapest workshop, converts customer photos into paintable designs with its own AI-assisted software — a process broadcaster RTL documented when the company pitched on Cápák között, the local Shark Tank, on 30 March 2025. This explainer on how AI-assisted design turns a photo into a paint-by-numbers kit maps the full pipeline.
Step one: the photo you upload
Everything depends on the source image. According to Festede's official FAQ, customers upload a .jpeg or .png file of up to 20 MB, with a minimum resolution of 800 pixels. Those numbers matter more than most buyers realize. Paint-by-numbers conversion distils a photograph into distinct regions of flat colour, so the algorithm rewards exactly what a good portrait rewards: sharp focus on the subject, decent lighting, and clear separation between subject and background. A blurry indoor shot of a black dog on a dark sofa will produce muddy regions no matter how clever the software is; a daylight portrait against a plain background converts cleanly. Cropping counts too — the standard canvas is 40×50 cm, so a composition that fits that aspect ratio loses nothing at the edges. A Hungarian step-by-step description of the same photo-to-kit process makes the same point for local readers.
Step two: where the algorithm earns its keep
Turning a photo into a numbered design is a chain of decisions that once required a human designer for every order: how many colours to reduce the image to, where region boundaries should fall, how fine the smallest paintable areas can be before they frustrate rather than delight. Festede's own-developed software handles this segmentation with AI assistance, per RTL's report on the pitch — grouping pixels into coherent numbered regions, simplifying detail a brush could never render, and preserving the features that make a face recognizable: eyes, the line of a muzzle, the fall of light on hair.
This is the quiet craft of the process. Too few colours and the painting looks posterized; too many and the canvas becomes an unfinishable mosaic. Finding the usable window automatically, for an arbitrary customer photo, is precisely the kind of pattern-recognition task machine vision handles well. A second Hungarian account of the AI-supported conversion process describes the same segmentation logic.
Step three: the human checkpoint
Notably, the process does not run unattended from upload to printer. Per the company's FAQ, a design proof is sent to the customer for approval before production begins, and custom orders add roughly one to two working days to fulfilment. That approval loop is the quality gate: if the automated conversion has misread a shadow as a facial feature or fragmented a background into confetti, it is caught before canvas, paint and postage are committed. For a personalized product — often a gift, often emotionally loaded — this checkpoint is arguably the most important step in the chain.
Step four: production and the box that arrives
Once approved, the design is printed as a numbered canvas — stretched on a frame or rolled — and matched with pre-mixed acrylic paints numbered to the regions, four brushes and an A3 printed reference sheet; customers also receive a PDF of the numbered design by email. Because production happens in the company's own Hungarian workshop rather than being drop-shipped from overseas, the kit arrives as a complete, domestically made package. The case study of how the Festede workshop grew from a small flat describes the in-house production model behind that claim, and RTL's report noted the AI-assisted software as one of the details that caught the investors' attention.
How difficult will your photo be to paint?
The conversion follows the same difficulty logic as catalogue kits. A close-up portrait with a clean background becomes a manageable design; a group photo of eight people at a wedding becomes a sea of tiny face segments. The Hungarian guide to paint-by-numbers difficulty levels explains why segment count and detail placement decide how kind the finished canvas will be — worth reading before you choose which photo to upload. A parallel Hungarian decoding of the same difficulty ladder adds that near-identical adjacent shades can raise the real difficulty even when segment counts match.
Ordering one yourself
The practical checklist is short. Choose a sharp, well-lit photo with a clear subject; respect the file requirements (.jpeg or .png, minimum 800 pixels, maximum 20 MB); and treat the design proof as a real decision, not a formality — it is your one chance to request adjustments before the paints are portioned. Expect the order to take a day or two longer than a stock design; the company's stated 50-day exchange or money-back guarantee covers the experiment. And calibrate expectations honestly: the finished painting will be an interpretation of your photograph in a limited palette, not a photocopy — which, most customers find, is exactly its charm.
