What are the best settings for scanning old photos for AI colorization?

Direct Answer

The best settings for scanning old photos depend on whether you are using a dedicated flatbed scanner or a smartphone app, but the core principle remains the same: capture the maximum amount of original detail so that an AI colorization tool like colorizethis.io has clean, high-resolution data to work with. For most users, a resolution of 300 DPI is the practical minimum for standard prints, while 600 DPI is recommended for small snapshots or photos with fine text and patterns. Scanning in 24-bit or 48-bit color depth, even if the original photo appears faded, preserves the subtle tonal gradations that AI models rely on when guessing and applying new color. The file format should be TIFF or PNG rather than JPEG, because lossy compression introduces artifacts that can confuse colorization algorithms and produce banding or false color in smooth areas like skies and skin. If you are working with negatives or slides, a dedicated film scanner at 2400 DPI or higher will yield far better results than a flatbed with a transparency unit, and the resulting raw scan will give colorizethis.io more luminance data to interpret.

Also worth reading: What are the definitive Topaz Video AI colorization settings guide for achieving professional results? · What are the best AI colorization tools in 2026 and how do they compare for restoring old photos? · How do I verify AI photo colorization accuracy for historical photos?

How and Why Scanning Settings Affect AI Colorization

AI colorization models are trained on millions of image pairs, and they learn to map brightness and texture patterns to plausible color palettes. When a scan is too low in resolution, the model lacks the fine-grained texture of grain, fabric weave, and facial features that guide realistic color placement. A scan saved as a JPEG at high compression may show blocky artifacts around edges, and the AI can misinterpret those blocks as color boundaries, leading to bleeding or halo effects in the output. Scanning in a higher bit depth captures more information per channel, which matters because old photos often have a very narrow dynamic range after decades of fading; the AI needs to see the full tonal structure to decide whether a shadow should be blue, brown, or black. The scanning process itself also introduces a type of noise called scanner noise, which is different from photographic grain, and AI models can sometimes treat uniform scanner noise as a texture to colorize, producing an unnatural speckled look. For these reasons, the settings you choose at the scanning stage directly shape the ceiling of what any colorization tool can achieve.

Practical Scanning Steps for Old Photos

Start by clean the photo gently with a soft, lint-free cloth to remove dust and loose particles, because scanner glass will magnify every speck into a dark spot in the final image. Place the photo face-down on the scanner bed, aligning it to the top-right corner or using the scanner's preview frame to ensure it is square and not skewed. If your scanner offers a digital ice or dust removal feature, use it with caution, as aggressive dust removal can erase fine details like hair strands or text on the back of a photo that might be historically valuable. Set the resolution to 600 DPI for standard prints and 1200 DPI for small wallet-size snapshots or slides, and choose 48-bit color depth if the option is available. Before committing to a full scan, run a test scan of a corner or a representative area and zoom in to 100 percent on your screen to check for focus issues, moiré patterns from printed half-tone dots, and color casts from the scanner's lamp. Once the test scan looks clean, scan the full image and save it as an uncompressed TIFF file, naming it with a descriptive prefix that includes the date and a sequence number so you can track which scans have been colorized.

Comparison of Scanning Methods

One of the most frequent mistakes is scanning at a low resolution to save time or disk space, then wondering why the colorized result looks blurry or smeared. A 150 DPI scan of a 4x6 inch print will produce a file that is only 600x900 pixels, which is too small for modern AI models that expect at least 1024 pixels on the longest edge to generate convincing color. Another common error is using the scanner's auto-color correction aggressively, which can clip shadows and highlights and remove the very tonal information the AI needs to make smart color decisions. Users also forget to clean the scanner glass between photos, and a faint smudge from a previous scan can transfer to the next image, creating a ghostly overlay that the AI may interpret as a color gradient. Scanning reflective photos at an angle instead of perfectly flat introduces parallax distortion, and while some software can correct for this, the resampling process softens the image and reduces the effective resolution. Finally, saving the scan as a JPEG with maximum compression is tempting because the files are small, but the lossy artifacts are exactly the kind of noise that confuses AI colorization and can produce strange color blotches in areas of fine detail.

When to Act and What to Expect

Old photos are subject to ongoing chemical decay, and the window for digitization before irreversible damage sets in is measured in years, not decades. Fading accelerates in humid environments, and a photo that has lost 20 percent of its contrast to UV exposure will still yield a usable scan for AI colorization, but one that has turned nearly white will provide almost no tonal data for the model to work with. If you have a collection of more than 100 photos, plan to scan them in batches of 20 to 30 per session to avoid fatigue and maintain consistent quality, and store the scanned TIFF files on a separate hard drive or cloud service with at least two redundant copies. The colorization process itself on a tool like colorizethis.io typically takes only a few seconds per image once the scan is uploaded, but you should budget time for reviewing the results and manually correcting any color errors, because AI colorization is a suggestion, not a restoration. Expect that very faded or heavily damaged photos will require more post-processing than fresh scans, and in some cases the AI may produce colors that are historically plausible but not strictly accurate, since the model cannot know the original shirt color or wall paint shade without additional context.

Cost and Pricing Considerations

Dedicated flatbed scanners suitable for photo digitization range from approximately $150 for entry-level models like the Epson Perfection V39 to $600 for higher-end units with infrared dust removal and faster scan speeds. Smartphone scanning apps are generally free, though some charge a one-time purchase fee of around $5 to $10 to remove watermarks or unlock higher-resolution export options. Professional scanning services charge anywhere from $0.50 to $2.00 per photo for standard prints, and $5 to $15 per negative or slide for high-resolution drum scanning, which is generally unnecessary unless you are working with archival-grade 35mm or medium format film. The colorization tool itself may offer a free tier with watermarked or lower-resolution output, and a subscription tier that unlocks higher-resolution downloads and batch processing for a monthly fee of roughly $10 to $30. When calculating the total cost of a digitization project, factor in the time required for scanning, the storage costs for keeping both raw scans and colorized versions, and the potential need for manual retouching if the AI produces artifacts that require correction in a photo editor.