How to
How to remove the background from a product photo (2026 guide)
Last updated: May 10, 2026
Every e-commerce listing starts with a clean product cutout. The bad news: most online background removers were built before 2024 and visibly fail on hair, fur, glass, jewelry chains, and reflective metals. The good news: you can now do this in 6 seconds with a 2026 model called BRIA RMBG 2.0. This guide walks through the practical steps and the technical why.
The 5-second version
Open palmou.com → sign up with Google → drop your product photo → don't pick a scene → click Generate → download the transparent PNG. That's it. The full version below explains why this works better than remove.bg or Photoroom for products with tricky surfaces.
Why most background removers fail on products
Tools like remove.bg, Photoroom's free tier, and Pixelcut use binary alpha masks. Each output pixel is either 100% kept or 100% removed. That's fine for a cardboard box on a white tablecloth — terrible for anything with soft edges.
The failure modes you'll see:
- Hair turns into hard halos. Strands disappear.
- Jewelry chains develop gaps where small links should be.
- Glass becomes opaque or develops weird stretched edges.
- Watch crystals and screen glares lose their reflectivity.
- Faux fur, knitwear fringe, and lace go from soft to chiseled.
The 2026 fix: 256-level alpha channels
BRIA RMBG 2.0 outputs a 256-level alpha channel. Each pixel carries a transparency value from 0 to 255 instead of just 0 or 1. The model is explicitly trained to capture partial transparency — chrome reflections that are 70% transparent, glass edges that are 40% transparent, hair strands that are 90% transparent at the tip and 10% at the root.
In practice this means: when you composite the cutout onto a new background, the soft edges blend correctly. No halos. No hard cuts. No cleanup pass in Photoshop.
Step-by-step on Palmou AI
1. Sign in with Google at palmou.com. You get 1 free credit on signup, no card.
2. Drop your photo into the editor. JPG, PNG, or WebP up to 10 MB. The editor previews it locally before you spend a credit.
3. Don't pick a scene. The scene picker is for full scene generation; for background removal alone, leave it blank.
4. Click Generate. BRIA RMBG 2.0 runs in about 6 seconds. The result loads in the editor as a before/after slider.
5. Click Download. The output is a WebP (or PNG when transparency is required) at your source's native resolution — a 4K input gives a 4K cutout.
When to use this vs full scene generation
If you only need the cutout — to drop into your own design template, your own composite scene, or a marketplace listing that already has a brand background — the background-removal-only flow is the right choice. It's faster (6s vs 15s) and preserves your source's full resolution.
If you want a finished catalog image (cutout + new background + lighting), pick a scene before clicking Generate. Lighting and shadows in the new scene adapt to your product, so the result looks shot, not pasted.
Common gotchas
- Phone photos with severe motion blur won't get cleaner. The cutout will preserve the blur.
- Source photos under 100 px on each side are rejected — too small for a usable result.
- If your background is the same color as your product (white shirt on white wall), the cutout may include some background noise. Use a contrasting backdrop when shooting.
- Glass-on-glass scenes (perfume bottles on glass shelves) are still hard. BRIA does the best public job but a manual touch-up may help on hero shots.
Try it free — 1 credit on signup, no card, no watermark.
Sign up with GoogleRelated guides
Amazon product photo requirements (2026): the complete checklist
Amazon rejects roughly 1 in 4 product images uploaded by new sellers — almost always for the same handful of policy violations. This guide is the definitive che…
Etsy listing photos: how to make handmade products look retail-ready
Etsy lets you upload up to 10 photos per listing — and the conversion difference between a 3-photo listing and a 10-photo listing is enormous. The bottleneck fo…