Topaz Image Colorization: change one locked still (8cr)

Topaz Image Colorization is an edit. Point at the object. A new text-to-image throws away the still you approved.

Topaz Labs, 8 credits, categories edit-image and image-to-image, requires_image: true. Description: "AI-powered colorization of black-and-white images with natural, realistic color restoration." Features: image_colorization, restoration, enhancement. No listed max resolution, no aspect enum, no qualities. Flat 8-credit call. audio does not apply. The "object" on this row is chroma. The geometry you already approved — the scan, the crop, the face — stays. A new generate will invent a different photograph that happens to be in colour.

Colorization is not a restyle generate

This SKU restores colour on a black-and-white (or faded) still. It is not "make it look like a Wes Anderson still." It is not text-to-image. If you needed a different pose, a different street, a different decade, you needed a generate. If you needed this scan to carry plausible colour so it can sit in a slideshow or become frame one of a motion job, you are here. 8 credits.

requires_image is true. There is no text-only colourization. Upload the plate. The catalog does not list a saturation enum on the model record; the colorize an old photo finishing route is the surrounding job. The agent phrasing is colorize an old black-and-white photo. AI photo editor is the shelf. There is no Topaz provider hub.

Colour is inferred, not historically verified. Treat the result as a restoration pass, not an archive fact. That limit lives on the finishing page; it is the honest brief for this SKU too.

Keep the scan. Change the chroma.

The still is the contract, same as LongCat, different delta. LongCat rewrites objects from a long instruction at 15 credits. Topaz rewrites colour at 8 credits. Using Topaz as a back door to a new image — uploading a colour photo and prompting a fantasy restyle — is the wrong label. Using a 2-credit generate to "recreate grandma in colour" is how you lose grandma.

No duration, no audio, no 1080p claim. If the next step is motion, colorize first, then image-to-video. Animating the black-and-white scan and hoping the video model invents period colour is how identity and chroma both drift.

Degraded, tiny scans give the model less to read. Upscale before colorize if the file is a postage stamp. That is a different row. This row is colour restoration on a still you would already keep in grayscale.

8 credits is an edit price. Spend it on a keeper.

If the scan is the wrong crop, colorizing it is an 8-credit prettier wrong crop. Crop, dust, and rotate first. Then colorize. Then, if the campaign needs motion, animate the colour plate.

A new t2i "in the style of an old photo, but colour" will not match the legal likeness, the wallpaper, or the date in the frame. Topaz exists so you do not have to fake the scan. Point at chroma. Leave the rest.

FAQ

Does Topaz Image Colorization require an image?

Yes. requires_image is true. The input is the black-and-white (or faded) still. No still, no restoration.

What am I actually changing?

Colour. The description is natural, realistic color restoration. Geometry, crop, and identity should hold. If you need those to change, you want a different edit or a generate, not this SKU.

Why not generate a colour version from a prompt?

Because a generate throws away the still you approved. 8 credits here is the catalog price to keep that still and add chroma. A new t2i is a new person.

Can I deliver this as video?

Not on this row. Image content type, 8 credits, no durations. Colorize, then send the colour still to an image-to-video model if the cut needs motion. Do not ask Topaz to animate.