Black and white to color, converted.

“Converter” normally promises a swap; this is not a swap. Most things called a converter change a container — JPEG to PNG, HEIC to JPEG — and MiniPx has one of those, running in your browser, doing exactly that. Some are filters instead: one arithmetic rule pushed across every pixel, blind to whether it is crossing a face or a hedge, which is why the “colorize” button in a photo editor yields a tinted wash rather than a photograph. This tool belongs to neither family. Nothing is swapped and no rule is applied uniformly. A trained network looks at your picture, decides what it is looking at, and writes a different answer into every part of the frame. Expect the behaviour of a judgement, not of a conversion — including the fact that two runs on the same file need not agree exactly.

There are no controls because there is nothing honest to control. People arriving from the word “converter” often look for a strength slider, a palette preset or a choice of eras. None of those exist here, and adding them would be theatre: the model makes one prediction per region, and a global warmth control would simply lay a cast over its work, which any editor already does better. What you can do instead is treat the output as a base layer. Colour the whole frame in one run, then correct the two or three regions whose real colour you actually know — a dress, a car, a football strip — which turns hours of manual colouring into minutes of correction. The reasoning behind that split is set out on the AI photo colorizer page.

One direction is lossless; the other is not. Desaturating a colour photo is arithmetic you can repeat forever with identical output. Going the other way is a guess, and that asymmetry has practical consequences. You can take the file this tool gives you, strip the colour in any editor, and land back on something that matches your input, because the light and dark were never altered. You cannot round-trip in reverse and expect stability: feed the desaturated result back in and the second answer may differ from the first. So if a photograph was originally in colour and somebody desaturated it, dig out that colour original instead — no prediction beats the data. And if the file needs more than colour, the repair models are separate on purpose: the face restorer for blurred faces, the object remover for a small mark over predictable surroundings.

What it costs, and what happens to your file. Because the model runs on our hardware rather than yours, this is one of the seven Cloud HD tools and it is part of MiniPx Pro: 150 runs a month on Pro Monthly ($3.24 the first month, then $4.99), 200 on Pro Annual ($29 the first year, then $39), prices excluding tax, resetting on the 1st with no rollover and a 30-runs-per-hour ceiling. A failed run is not charged. Your photograph is uploaded for that run, encrypted in transit, processed in the United States, deleted when the run finishes and in all cases within 24 hours, and never used to train anything. Everything else on MiniPx — including the format converter people often came looking for — runs entirely inside your own browser. Pricing detail lives on the pricing page.