Recovering when the image is wrong

Sometimes you can't re-source. Old design assets. Customer-supplied images. Stock images you bought. AI-generated outputs you can't easily regenerate (lost the seed, used up your credits, the source tool changed). The image is what it is.

This article covers the tactics for getting clean cutouts from imperfect sources, using ReadyPixl tools.

Diagnose first: what's wrong with the image?

Before reaching for fixes, identify which problem you have:

Each problem has a different fix below.

Problem 1: Background isn't a solid color

The image has a textured, patterned, or gradient background.

Fixes:

1. Use Color Removal in Manual mode — click the color picker on the dominant background color (not Auto, which only catches the average)
2. Add a second color tab in Color Removal for the next dominant background color
3. Add a third color tab if needed (max 3)
4. Raise Tolerance to 50-60 to catch more variation per color
5. If still messy, the source image is genuinely wrong for Color Removal — you may need AI Background Removal (15 credits per use)

Problem 2: Background color appears in the subject

The image is a white subject on a white background. Or a black logo on a black background. Color Removal can't tell them apart.

Fixes:

1. Turn ON Contiguous in Color Removal. This only removes background pixels connected to the edge of the image — interior matching pixels (your logo, your accent, your white packaging text) stay intact.
2. If Contiguous doesn't help (the background and subject are connected via thin areas), the image genuinely needs to be re-sourced or done by hand in another tool.

Problem 3: Multiple background colors

The image has a gradient or two distinctly different background regions.

Fixes:

1. Use Color Removal's color tabs — add up to 3 colors, each with its own Tolerance. Pick each background color manually with the color picker.
2. For a gradient, raise Tolerance significantly (50-70) on the picked color so it catches the full range
3. For very complex backgrounds (busy scenes, multi-color patterns), Color Removal isn't the right tool — Use AI Background Removal

Problem 4: Soft / blurry edges

Cutout looks chopped, jagged, or the edges look hard and pixelated.

Fixes:

1. Set Edge Feather to 2-3 in Color Removal — softens the cut edge
2. For very high-detail edges (hair, fur, fringe), Edge Feather 3-5 works
3. Don't go above 5 — too much feathering looks blurry

Problem 5: Halos around the subject

A faint ring of color is visible after Color Removal — most obvious when you put the cut subject on a contrasting background.

Fixes (in order):

1. Add Transparency Cleaner after Color Removal. Set its slider to 30-50. Targets exactly this issue.
2. Set Edge Trim to 1 or 2 in Color Removal — shaves the edge pixels off
3. Raise Color Removal Tolerance by 5-10 — the halo is partly because Tolerance was too low to catch the edge gradient

Problem 6: Stray pixels left in the cutout

Tiny dots of background color remain after Color Removal, scattered around the subject.

Fixes:

1. Add Speckle Remover after Color Removal — defaults (Max cluster size 50) catch most stray pixels
2. Use View mode first to see what would be removed before switching to Delete
3. Raise Max cluster size to 100 or 200 if leftover bits are bigger than the default catches

Problem 7: Subject too small in the frame

The subject is centered but takes up less than ~30% of the frame, leaving a lot of background to deal with.

Fixes:

1. Add Trim to your pipeline — cuts off the empty edges after Color Removal
2. Then Reposition to put the trimmed subject on your target canvas size
3. For a very small subject (under 10% of frame), the image's actual resolution is probably too low for upload-ready output — consider regenerating at higher resolution / re-sourcing

The full salvage pipeline

For a maximally bad source image, this pipeline gives the best Color Removal result possible:

This is the "kitchen sink" approach — every cleanup tool ReadyPixl has. For a difficult source, it produces the cleanest possible result Color Removal can manage.

If the result is still not clean enough after this pipeline:

• Use AI Background Removal (15 credits per use, ~1.5¢ each) — handles images that Color Removal can't
• Or do it by hand in another image editor (Photoshop, GIMP, Photopea)

When AI Background Removal is the right call

For these image types, don't waste time trying to make Color Removal work — Use AI Background Removal:

• Images of people (hair is too complex for color-based removal)
• Furry subjects (pets, plush toys)
• Transparent or semi-transparent subjects (glass, water, smoke)
• Subjects against complex / busy backgrounds (cityscapes, nature)
• Anything where the subject and background colors heavily overlap

Color Removal is for solid-color backgrounds. AI Background Removal is for everything else.

When the image is genuinely unsalvageable

Some images can't be cleaned up to a usable cutout no matter what tool you use:

• Out of focus / blurry source — nothing recovers focus that wasn't captured
• Very low resolution (under 500 px on the long side) — too few pixels to work with
• Severely overexposed (subject blown out white) — no detail left to keep
• Severely underexposed (subject crushed black) — same in reverse

For these, regenerate or re-source. No tool restores information that wasn't there.

A note on AI-generated source images

If your "bad image" is from Midjourney, ChatGPT, Leonardo, or another AI generator, regenerate with a better prompt before fighting with cleanup.

The key prompt addition: "isolated on solid color background" or "on a flat black background" or similar. AI tools generate much cleaner backgrounds when you ask for them.

See the AI image cleanup workflow and Midjourney cleanup for the prompting + pipeline approach.

What to read next

• The 3 rules for the cleanest background removal — for the next time you DO control the source
• Prompting AI for clean backgrounds — for AI-generated sources
• What to avoid in your background — what makes Color Removal struggle
• Color Removal — full reference for the tool
• Speckle Remover — for stray pixels
• Transparency Cleaner — for halos
• Quality issues — broader troubleshooting