A focused AI photo colorizer that turns black-and-white images into realistic color using a two-model pipeline: one model generates a caption describing the image, and the second applies color based on that text . Its key differentiator is control—instead of a single automatic result, it offers 20 named filters and a text-prompt editor that lets you correct misidentified objects or steer the overall palette (e.g., “warm skin tones” or “1920s street scene”) . It was created by Emil Wallner, a machine learning researcher and former Google Arts & Culture resident artist . Pricing is freemium: the free tier produces low-resolution, watermarked previews; paid plans start at $39/month (Pro, 8,000 credits) with a Basic free tier of 200 credits, and high-resolution downloads cost 10–20 credits depending on size . A video colorization tier also exists (Ultra $89/mo, Mega $189/mo) . Critical caveats: The free tier is explicitly a preview, not a usable output—watermarked and low-resolution . The tool is single-purpose: it does not offer general photo editing, retouching, or object removal, and results degrade on badly damaged or ambiguous source images . Independent review volume is thin, and the product’s pricing has shifted from its early “completely free” state in 2022 to the current credit-based model . And while it handles portraits well, it can misidentify objects and apply incorrect colors, which is why the text-prompt correction step exists .

Pros & Cons:

Pros: The text-prompt control is the core differentiator—instead of accepting whatever the AI guesses, you can edit the caption it generated, correct misidentified objects, and re-run colorization until the result is closer to the original or to your intended mood . The 20 named filters give quick alternative palettes without any prompt writing, and the “Surprise Me” option generates prompts for you . Results on clear portraits and archival family photos are consistently praised, with users comparing it favorably to older colorizers like ImageColorizer . The upload-and-download flow is genuinely simple—no signup required for the free preview, and no steep learning curve . It supports high-resolution exports (up to several thousand pixels) and many file formats on paid plans, which matters for print and archival use . A Twitter bot (@palettefm_bot) also lets you colorize by mentioning it in a post .

Cons: The free tier is not usable for real work—outputs are watermarked and low-resolution, so evaluation of final quality requires paying . It is a single-purpose tool: no retouching, no object removal, no general editing, and results degrade on damaged or ambiguous sources, meaning archival work often needs a separate retouching step . Object misidentification is a documented limitation—the AI can fail to recognize items (e.g., raspberries, oranges) and apply wrong colors, requiring manual prompt correction . The pricing model is credit-based and can feel restrictive, and subscription credits expire if cancelled . There is no mobile app, no offline mode, and no true self-hosted option . And the product’s early “completely free” positioning has changed to a paid credit system, which may surprise users relying on older reviews .

Who is it for?:Genealogy hobbyists, family historians, and individuals restoring old family photos who want believable color on portraits and clear archival images without learning a full editor . Archivists, museums, and documentary teams that need controlled colorization with text-prompt correction and high-resolution exports . Users who want to experiment with different palettes and moods on a single black-and-white image, rather than accept one automatic result . Not suited for anyone needing a free tool for usable output (free tier is watermarked preview only), users requiring general photo editing or restoration of badly damaged images (it is a colorizer, not a retoucher), teams that need a mobile app or offline processing, or organizations requiring extensive independent third-party reviews before adoption.