DEV Community

Cover image for The 'Free' remove.bg Replacement Costs $144 a Year. Do the Maths First.
AheadKit
AheadKit

Posted on

The 'Free' remove.bg Replacement Costs $144 a Year. Do the Maths First.

Three days ago I wrote that remove.bg dies on December 1, and that the free tool on your own machine could replace it. The piece did well. Then deanlee left a comment I have been thinking about ever since. He asked the question I had dodged in the original piece: at what point does paying for Canva actually cost less than paying per removal somewhere else?

I dodged it with a soft sentence. Something about the subscription maybe being the saner choice for heavy users. That is not an answer. It is a shrug dressed up as advice. So I spent this morning doing the sums properly, re-weighing the files from that first test, and recounting the pixels behind a number I now regret printing. Here is the honest version.

Problem 1: the subscription costs $144 a year, and the break-even is 12 removals a month

Canva Pro is $18 a month if you pay monthly, or $144 a year if you pay once. That works out to $12 a month. One person only. Those are the US prices, read from Canva's own pricing page in October 2026 — Canva charges different amounts in other countries, so if your pricing page shows a different currency, run the same sums with your own numbers. The background remover sits inside that paid plan, and the free plan does not have it.

Now the other side. After December 1, the thing you compare it against is not remove.bg's $9 subscription. That plan stops existing. The honest comparison is paying per image, and remove.bg's smallest one-off pack is $3 for 3 credits. One credit takes the background out of one photo. That is $1 a removal.

Twelve dollars divided by one dollar is twelve. If you pay Canva yearly, you need to remove 12 backgrounds a month before the subscription beats paying a dollar a time. Pay monthly instead, and the line moves to 18 a month.

Bigger packs bend the number. The $49 pack works out to 65 cents a removal, and against that Canva only wins past 19 removals a month on yearly billing, 28 on monthly. The $199 pack drops it to 40 cents, and then you need 30 a month yearly, 45 monthly. So the real answer is a range, not a single number: somewhere between 12 and 30 a month, depending on how you buy the one-off credits.

And one more honest wrinkle. If you already pay for Canva because you use its templates or its brand tools, the remover costs you nothing extra. Your break-even is zero. This maths only bites if background removal is the thing pushing you to pay in the first place.

One gap I could not close: several reviews this year say Canva's paid AI tools share a monthly allowance, and I could not confirm whether the background remover draws from it. If you cut out hundreds of photos a month, look at the limit on your own plan before you assume it is bottomless.

Problem 2: the cheap price you remember is a ghost

This is the part that changes the answer most. In my first piece I quoted remove.bg's $9 plan, 40 images, about 23 cents each. That number is real, and it is also dead in two weeks. Subscriptions end on December 1 and unused one-off credits expire the same day. Comparing Canva against 23 cents is comparing against a price nobody will be able to pay after the shutdown.

The fair comparison is what you can actually buy on December 2: a dollar a removal at the small end. That is the number the break-even above uses.

If you still hold remove.bg credits, use them before December 1. After that they are gone, and so is the argument that paid removal costs 23 cents.

Problem 3: the PNG weighs more than I told you

In the first piece I said the shoe went in as a 128 KB JPEG and came out as a 632 KB PNG, and I waved at compression as the fix. I should have weighed it properly then. I have now, using the same cut-outs from that morning.

Same shoe, same pixels, three formats. The PNG is 632 KB. Saved as WebP with the transparency kept, at a quality setting where I could not spot a difference at normal viewing size, it is 94 KB. That is 6.75 times smaller. Flattened onto white and saved as a JPEG, it is 99 KB, or 6.37 times smaller. The portrait is worse: a 1.56 MB PNG becomes a 192 KB WebP, 8.13 times smaller, or a 234 KB JPEG, 6.69 times smaller.

So the "five times" figure floating around my draft notes was wrong in the generous direction. It is six to eight times on real photos with a real subject. The one exception was the mountain lake photo, where the PNG was only 3.5 times the WebP, because there was almost nothing in the file to compress in the first place.

I also built a small test page with one shoe on it, in all three formats, and measured what actually travels down the wire. The PNG page moves 632 KB. The WebP page moves 94 KB. Scale that to a shop with 40 product shots and the PNG version ships 24.7 MB of images where the WebP version ships 3.7 MB. That is the difference between a page that opens and a page your customer stares at.

The fix costs nothing. Convert the cut-out to WebP before it goes on a page, and keep the PNG only as your master copy. Lossless WebP barely helps, about 1.4 times smaller, so it has to be the lossy setting to matter.

Problem 4: the 71.7% number was right, and it meant nothing

This is the one I owe readers a correction on. In the first piece I printed "71.7% transparent" for the mountain lake photo, then said in the next paragraph that the number meant nothing because the remaining 28% was a grey blob. Both halves were true, and printing the number at all was still a mistake, because a table makes any number look measured.

I recounted it pixel by pixel from the actual file. 689,221 of 961,200 pixels are fully transparent. That is 71.70%. The number was correct. The shoe was 73.48% against the 73.5% I printed, and the portrait 66.78% against 66.8%. All three check out.

But here is what the recount found that the table hid. In the lake photo, only 7 pixels in the entire image are fully solid. Seven. The other 28.3% of the picture is made of half-transparent pixels, a thin wash of grey where the tool could not decide what was subject and what was sky. A transparency percentage counts pixels. It cannot tell you whether the pixels it kept form a shoe or a smudge. No percentage can. Only looking at the picture can, and when I looked, the picture was a ghost.

The recount also explains the file sizes above. Even the good cut-outs are mostly half-transparent edge pixels, 24% of the shoe and 32% of the portrait. Soft edges are why the tool looks good, and soft edges are exactly what makes a PNG balloon.

What I would actually do

If you remove backgrounds a few times a month, do not buy a subscription for it. Pay a dollar when you need one, or run the free tool on your own machine and spend eight seconds a photo instead. Twelve removals a month is the line, and most people never cross it.

If you cross it every month, and you do it inside Canva already, pay yearly and stop thinking about it. The $144 works out to $12 a month, and the per-removal price stops mattering.

And whatever you switch to, do not ship the PNG. Convert it. The cut-out is free. The page weight is the bill you pay later, every time someone opens your site.

Now I want the other side of this. How many backgrounds do you actually remove in a month, the real count, not the guess? And if you already switched away from remove.bg, what did you switch to, and what did it cost once you counted everything? If enough of you answer with real numbers, I will run the same maths on your answers and print the correction if I got any of this wrong.


P.S. Everything I write stays free — that part doesn't change. But if this one was worth something to you, you can buy me a coffee ☕ (one-time or monthly, both work). It covers the API credits and hosting behind these experiments — and honestly, it's also the nudge that gets the next one started on the evenings I'd otherwise skip it. — Salam

Buy me a coffee

AheadKit on Gumroad — Beginner-Friendly Books & Packs

Top comments (0)