Background remover PNG too big? One cutout went...

Background remover PNG too big? One cutout went from almost 7 MB to 356 KB as WebP

By SJ
6 min read
Bar chart of remove.bg's sample cutout in bytes: PNG 6,731,926, Smart PNG 2,407,602, Smart WebP 356,282, Smart AVIF 194,550
On this page

The sample cutout that remove.bg publishes on its API page is a 6,731,926-byte PNG. Through Iminify at Smart, its default lossy level, the same file came back as a 356,282-byte WebP (356 KB) with every transparency value unchanged. The AVIF was 194,550 bytes and changed transparency values by up to 29 of 255; the PNG was 2,407,602 bytes and kept 32 of the 256 values.

So save the cutout as WebP, which changed no transparency value, and keep the PNG from the remover as your master. PNG to WebP does it. Below is what each choice gives up.

Most of a cutout PNG is the picture, not the transparency

Every pixel of a cutout carries a number from 0 to 255 for how see-through it is, on top of its colours. That is the transparency channel (the alpha channel, in editors). Saved on its own as a grey PNG, the sample's channel is 126,988 bytes. Inside the PNG it costs more: after a lossless repack (the size Iminify's Lossless PNG gives, below) the file is 6,147,485 bytes, and without the channel it is 5,447,020, so the channel accounts for 700,465 bytes, 11.4%.

On three cutouts we made, from the photographs of our first post, it accounts for 16.7% to 19.4%. Our PNG post never measured that on a real cutout. The picture is the rest.

About 63% of the sample's pixels are fully transparent, and every one of them holds the same black, so no second picture is hiding under them.

remove.bg's own table lists its PNG at 7 MB, its WebP at 6 MB and a ZIP at 3 MB (the sample files weigh 6,731,926, 5,885,026 and 2,489,246 bytes). The ZIP is a JPG plus a separate transparency PNG that you put back together yourself. The sample PNG has 9,996,213 pixels and the WebP 11,107,122, so per pixel the WebP is 21.3% smaller than the PNG. That WebP is lossless, and nowhere near 356,282 bytes. The page says background removal moves to Leonardo.Ai on 1 December 2026.

What each format did to the same cutout

We uploaded the sample to Iminify through its API on 7 November 2026 and asked for the levels and formats below. The scores are SSIMULACRA 2 (100 is identical), measured by us on the files we got back.

Iminify result Bytes Smaller than the 6,731,926-byte PNG Score Transparency values
Lossless PNG 6,147,485 8.7% 100 all 256 kept
Smart PNG 2,407,602 64.2% 70.11 32 of 256 left
Lossless WebP 3,509,220 47.9% 100 all 256 kept
Smart WebP 356,282 94.7% 70.85 256, none changed
Smart AVIF 194,550 97.1% 70.08 256, largest change 29

Smart is lossy: it changes the colours to save bytes, which is what a score under 100 measures. A second upload of the same file gave byte-identical results. Our three cutouts ordered the same way. Smart WebP was 87.3% to 91.2% smaller than the PNG, Smart AVIF 97.7% to 98.2% and Smart PNG 62.2% to 70.7%.

Compress PNG keeps a PNG but leaves fewer transparency values

Smart and Ultra, its smaller-file level, give the PNG a palette of at most 256 entries, and each entry has to cover a colour and a transparency together. The sample came back with 32 distinct transparency values; ours (cup, raspberries, canoe) kept 56, 64 and 62 of the 248 values they had. The largest change to one pixel's transparency was 58 on the sample and 30 to 42 on ours, out of 255. Smart WebP changed none.

If a form accepts only PNG, use Compress PNG at Lossless and expect little: it saved 8.7% on the sample and 0.1% to 1.5% on ours.

AVIF is smaller partly because it changes the transparency too

Smart AVIF's largest change to a transparency value was 29 on the sample and 52, 36 and 41 on ours, in the same order. That's the default post 5 found: avifenc without --qalpha codes the transparency at lossy quality too.

We encoded all four cutouts at -q 70 both ways. The lossy transparency made the file 13.3% smaller on the sample and 48.0% to 68.5% smaller on ours, and the score moved by 0.02 to 0.41, in favour of the exact transparency. Over a white background the mean worst-channel error across the partly transparent pixels differed by 0.1 to 0.7 out of 255 between the two. The lossy transparency costs little, so AVIF is a fair choice when the file must be as small as possible.

Below, one stretch of hair edge from our canoe cutout at 4x on blue: the cutout, Smart PNG, Smart WebP and Smart AVIF. The outline holds in all three. Speckle appears in the PNG, while the WebP and AVIF smooth the hair texture. (The blog saves this strip as WebP too, so it is a little softer than the files.)

Four enlarged crops of the same hair edge on blue: the cutout, Smart PNG, Smart WebP and Smart AVIF

How to shrink a cutout

  1. Keep the PNG the remover gave you. A lossy file can't give the edge back.
  2. Resize to the width the page shows. The sample re-encoded as WebP at quality 80 (our run, not Iminify's) was 554,312 bytes at 2,721 pixels wide, 299,324 at 2,000 and 137,282 at 1,200. Trimming the empty margin saved 1.3% of the PNG and 0.3% of the WebP, so skip it.
  3. Open PNG to WebP and leave it on Smart. Iminify can't remove backgrounds; it shrinks the file you already have.
  4. Lay the result on a coloured background and look along the edge before you upload it.
  5. For a big file on a page you control, try PNG to AVIF. Auto picks the smallest format only up to 8 megapixels and goes straight to WebP above that, which is why the sample (10 megapixels) came back as WebP and not the AVIF that was 45.4% smaller. Our cutouts, at 2.2 megapixels, came back as AVIF.

How we measured

The files are remove.bg's sample set from its API page (fetched 7 November 2026; we show none of its pixels) and three cutouts we made with rembg 2.0.87, model isnet-general-use, from the Unsplash-licensed photographs of our first post: a coffee cup by Justin Leibow, raspberries by Ben Moore and a canoe by Roberto Nickson. In our three cutouts rembg's mask reached 255 on under 1% of the subject's pixels, so we set every transparency value of 247 or more to 255.

Scores come from ssimulacra2 (libjxl 0.12.0) on the RGBA files, decoded with dwebp and avifdec; it blends transparency against dark and bright backgrounds and keeps the worse score, as in our AVIF and WebP post. The other runs are ours. On the sample's full-size WebP decoded to PNG (2,721 pixels wide) we ran cwebp 1.5.0 with -m 6 -pass 10 -mt (adding -q 80 for the sizes, after an ImageMagick resize for the narrower ones). avifenc 1.2.1 with -s 4 -y 420 ran at -q 70, with and without --qalpha 100, on that file and on the three cutouts' own PNGs. optipng 0.7.8 with -o2 -strip all repacked the PNGs behind the channel figures.

We uploaded nothing to remove.bg, so what its app exports is untested.

Questions this post answers

Share this post

Share on X
Share on Facebook
Share on LinkedIn
Share on Reddit
Share on Hacker News
Email
Copy link
Link copied
The same photo of a canoeist twice: upright on the left under the label "tag read", and lying on its side on the right under the label "tag gone".

Why does my photo upload sideways? The rotation lives in a tag, and stripping metadata throws it away

We stored a photo on its side with the EXIF tag Orientation 6 and opened it in Chrome 156. It stood upright. Then exiftool -all= removed the metadata, and the photo lay on its side. So did the copies saved by magick -strip and jpegtran -copy none. That is how a photo can look right on your phone and arrive sideways on a website. The file holds the pixels the way the camera sensor wrote them, plus a tag that tells a viewer how to turn them. A site that ignores the tag, or software in its pipeline that drops it, shows the pixels as stored. The fix is to turn the pixels themselves so no tag is needed, and the steps are below. Removing metadata doesn't do that.

7 min read
Blue and violet gradient cover with the white text Does Shopify compress images? Chrome got a median 11% of the source, a field of faint squares on the right and the Iminify mark.

Does Shopify compress images? Yes, when it serves them: Chrome got a median 11% of the source, 25% at full size

Yes, in the files it serves, and not by one fixed amount. On 9 November 2026 we loaded 78 pages from 43 public Shopify stores in Chrome 156 and asked Shopify's image servers for 459 of the images on them. The file Chrome got was a median 11% of the size Shopify reports for the source. At full size, the source's own pixel size instead of the width the page asked for, it was 25%. Shopify's help says images "aren't compressed when you upload them to the Shopify admin". Its CDN page says "File size is automatically optimized when you upload to Shopify". We uploaded nothing, so this is about what Shopify sends, not what it keeps.

7 min read
Three crops of the smallest text on the Inter sheet with the fields Gemini read: the PNG (228 KB) 32 of 32, JPEG quality 40 (172 KB) 32 of 32, JPEG quality 5 (76 KB) 9 to 14 of 32, where the digits are visibly smeared.

Does compressing an image change what Gemini reads? Full-colour JPEG quality 95 to 40 didn't, shrinking to 600 pixels did

Gemini 3.8 Flash read the same number of fields from a full-colour jpegli JPEG saved at quality 40 as from one saved at quality 95, on both of the text sheets we tested. A copy shrunk to 600 pixels wide, a PNG with no JPEG loss in it, read 15 fields fewer than the quality 95 file on both, comparing medians. A field is one code, price, date or 12-digit number, and each sheet holds 192 of them as random strings, so there's nothing to guess from. Anthropic's vision page says "heavy JPEG compression can make text difficult to read" and gives no number. Google's image page advises "Use clear, non-blurry images" and doesn't mention compression. We read both pages on 8 November 2026. These numbers are for Gemini 3.8 Flash, through the API, on 8 November 2026.

7 min read

Try it on your own images

Drop a photograph in, pick Smart, and compare the result with the original before you download. Free, no account needed.