Colorize photos without upload

Turn old black-and-white photos into colour pictures – family photos, portraits, street scenes. The AI (DDColor) suggests fitting colours for sky, skin, plants, clothing and objects; the sharpness of your photo is kept at full resolution. Model and computation run on your device, nothing is uploaded.

🔒 No upload – your photos never leave your device
    100% = as suggested by the model. Less looks like a hand-tinted photo, more is bolder.
    coolerwarmer
    Colour by click Do you know what colour something was? Pick the colour, press “Set colour” and tap the spot – the AI finds the area. The first time the selection model loads (40 MB, from mytoolio).

      DDColor (Apache 2.0) · one-time 67 MB, served by mytoolio

      Pick a black-and-white photo on the left – it is colourised automatically.

      BeforeAfter
      Ready

      How it works

      1

      Choose photos

      Drop one or more photos on the field, pick them, or paste an image from the clipboard. The first time, the browser downloads the model (67 MB).

      2

      Colourise automatically

      Every photo is colourised within seconds. Compare before and after with the slider and make the colours softer or bolder with “Colour strength”.

      3

      Save

      Download the result as JPG, PNG or WebP – one by one or all together as a ZIP – or copy it to the clipboard.

      Frequently asked questions

      Are my photos uploaded?+

      No. The photos are colourised in your browser and do not leave your device. The AI model and the runtime are served by mytoolio itself – no other provider is involved.

      Are the colours real?+

      No – they are estimated. The model has learned from millions of colour photos what sky, skin, grass, wood or stone usually look like and applies that to your picture. It cannot know whether a dress really was blue or green. The result is a plausible interpretation, not a historical record.

      Which photos work well?+

      Photos with clearly recognisable subjects work best: people, landscapes, buildings, animals. Very dark, blurry or badly damaged pictures stay pale or blotchy. Sepia photos work too – the brown tint is replaced by the new colours.

      Is the resolution preserved?+

      Yes. The model only computes the colour (at 512 × 512 points) and lays it over the brightness of your photo at full size. Detail and sharpness stay unchanged; photos larger than 6,000 pixels on the long side (4,096 on phones) are reduced to that.

      Can I influence the colours?+

      The “Colour strength” slider controls how strong the colours are – from a delicate, hand-tinted look to bold tones. With “Colour by click” you set individual colours: pick a colour, tap the spot – the AI finds the whole area (dress, sky, car) and colours it. “Colour temperature” makes the whole picture warmer or cooler, “Automatic contrast” helps with flat scans. Tip: a colour photo whose colours have faded can be re-coloured as well.

      How long does it take, and does it work offline?+

      With graphics-card support (WebGPU in Chrome and Edge) a photo takes less than a second, otherwise about 10 to 20 seconds. After the first download the model stays in the browser cache; then the tool also works without an internet connection.

      Can the tool remove scratches and spots?+

      Yes – with “Remove scratches and spots” thin bright and dark scratches, dust and small spots are touched up before colourising. This needs no AI model and takes a few seconds: the tool looks for narrow marks that stand out clearly from their surroundings and fills them from the neighbourhood. “Strength” sets how much it touches. Fine hair, eyelashes or writing may be caught too – check with the before/after slider and use a lower strength if in doubt. Wide tears, creases and large stains stay; the object eraser suits those.

      Colourisation with DDColor-tiny (Xiaoyang Kang et al., Apache 2.0 licence) on onnxruntime-web – entirely on your device. The colours are an estimate by the AI and do not reproduce the actual colours of the scene.