Midjourney cleanup: prep AI designs for print or upload

End-to-end recipe for cleaning up Midjourney designs — strip the AI background, fit them to the canvas you need, get them upload-ready in one batch.

Midjourney generates beautiful designs. It generates them at sizes and with backgrounds that don't quite work for what you want to do with them.

This article walks the recipe for cleaning up a folder of Midjourney downloads and getting them into a usable, upload-ready batch.

What Midjourney gives you (and the gap)

Default Midjourney output:

Size: typically 1024 × 1024 pixels
Background: whatever the AI generated (busy scenes, soft gradients, abstract patterns)
Aspect ratio: square, unless you specified otherwise with --ar
File type: PNG

The gap between what Midjourney gives you and what print-on-demand sites / Etsy / your own designs need:

You need	Midjourney gives	Bridge
Large images (4500+ px)	1024 px	AI Upscale (10 credits per image) — pick 4x to hit 4500 px from a 1024 source
Transparent backgrounds	AI-generated scenes	Color Removal
Specific aspect ratios	Square (default)	Reposition
Consistent batches	Each design unique	Pipeline + presets

The pipeline

Step	Tool	What it does
1	Color Removal	Removes the AI-generated background
2	Trim	Cuts off the empty edges
3	Speckle Remover	Cleans up tiny stray AI pixels
4	Transparency Cleaner	Removes faint halos AI background removal sometimes leaves
5	Reposition	Centers the design on your target canvas (Amazon Merch, Etsy, etc.)

Step-by-step

Generate in Midjourney. Download the outputs to a folder on your computer.

Open ReadyPixl. Drag the Midjourney folder onto the editor.

Add Color Removal. Midjourney backgrounds are often gradients or busy scenes — defaults may not be enough. Try:

Tolerance: start at 40 (higher than usual since Midjourney backgrounds vary more in color than typical images)
Edge Feather: 2-3 (Midjourney subjects often have soft AI edges; feathering keeps them looking natural)
For AI-generated batches, the better long-term move is to prompt the AI tool to put your subject on a solid color background to begin with — then Color Removal works on autopilot at default settings. See Background Removal Tips for the technique.

Add Trim. Defaults work.

Add Speckle Remover. Midjourney sometimes leaves tiny stray pixels around the subject. Default Max Cluster Size of 50 catches most. Use View mode first to confirm.

Add Transparency Cleaner. Midjourney subjects can have faint semi-transparent halos. Set its slider to 30-40 and the halos disappear.

Add Reposition for your target spec:

Amazon Merch shirts: 4500 × 5400 @ 300 DPI, center-center
Etsy listings: 2000 × 2000 @ 72 DPI, center-center
Phone wallpaper: 1170 × 2532 (iPhone 13/14) @ 72 DPI
Social media post: 2048 × 2048 @ 72 DPI

Click Download All. Open the zip — Midjourney designs are now upload-ready.

Save as a preset named "Midjourney to Merch" (or whatever your target is). Future Midjourney batches go through in 30 seconds.

⚠️ The size gap (until AI Upscale ships)

Midjourney's 1024 × 1024 outputs are smaller than print-on-demand sites want. Two ways to handle this until ReadyPixl's AI Upscale premium tool ships:

Option A: Reposition without upscaling. The cleanest approach — your 1024 × 1024 design sits centered on a 4500 × 5400 canvas. The design will be smaller relative to the canvas (about 25% as wide). Most POD platforms accept this and show the design at print size.

Option B: Use Midjourney's built-in upscale. Midjourney's "Upscale (4x)" or "Vary (Strong)" buttons can produce larger outputs (2048 × 2048 or 4096 × 4096 in some modes). Use those before downloading. ReadyPixl then has bigger source images to work with.

When AI Upscale ships in ReadyPixl (paid, 10 credits per use — about 1¢ per image at 1,000 credits per $1), the workflow gets a step 8.5: AI Upscale to 4×, then Reposition to your target canvas.

Tuning per design type

Midjourney designs vary widely. Adjust the pipeline based on what you generated:

Design type	Pipeline tweak
Sharp, hard-edged subject (logos, geometric designs)	Color Removal Tolerance 30; Edge Feather 0
Soft, painterly subject (illustrated characters, photorealistic)	Color Removal Tolerance 40; Edge Feather 3
Subject with text in the design	Don't run Color Removal between text strokes — turn ON Contiguous
Multiple subjects (prompt with characters + objects)	Crop closer to the main subject before the pipeline (in Midjourney)
Very clean, transparent-friendly subject (already on near-solid background)	Standard pipeline; Color Removal handles it cleanly

Tips

Generate 3-5 variants per concept. Midjourney quality varies wildly. Generating multiple versions and picking the best after pipeline-running is faster than trying to rescue a bad output.
Write Midjourney prompts with cleanup in mind. Prompts ending with "isolated on white background" or "on a solid color background" make Color Removal much easier.
Save originals. Your ReadyPixl outputs are the deliverables. Your raw Midjourney files are the seeds. If a future cleanup pass is needed (when AI Upscale ships, or after Color Removal improves), you'll need the originals.
Save one preset per target. "Midjourney → Merch shirts," "Midjourney → Etsy stickers," etc. Switching between targets becomes one click.

What Midjourney + ReadyPixl can't fix

Bad anatomy (broken hands, warped faces, AI artifacts) — pipeline can't restore content that isn't there. Regenerate the bad ones in Midjourney.
Wrong aspect ratio prompts — if you generated square but need portrait, Midjourney needs to redo with the right --ar flag. Reposition can fit a square design on a portrait canvas, but the design will be small.
Misspelled text — Midjourney's text rendering is often bad. Pipeline can't fix the words. Regenerate with text removed and add text yourself in another tool, or use Watermark Text in ReadyPixl.