DEV Community

Cover image for APIC vs Photopea vs Pixlr vs Canva: The Ultimate Browser Photo Editor Showdown (11 Rounds, No Mercy)
Akhouri Anmol Kumar
Akhouri Anmol Kumar

Posted on

APIC vs Photopea vs Pixlr vs Canva: The Ultimate Browser Photo Editor Showdown (11 Rounds, No Mercy)

#showdev #webdev #productivity #ai

Description: "11 rounds, 6 fighters: privacy, AI, colour science, price, formats and mobile. A brutally honest battle of free browser photo editors, by the guy who built one of them."

๐ŸฅŠ APIC vs Photopea vs Pixlr vs Canva: The Showdown

Ladies and gentlemen... in the blue corner, the PSD monster. In the orange corner, the AI credit machine. In the purple corner, the design giant. And in the green corner, a one-person indie editor that never uploads your photos... APIC.

Tech YouTubers have flagship-phone shootouts. Browser photo editors never get the same treatment, even though millions of us open one every week to crop a screenshot, cut out a background or squeeze a hero image under 100 KB.

So let's give them the treatment. 11 rounds. 4 main fighters. 2 special guests. One scoreboard.

โš ๏ธ Fair-play disclosure (read this first)

I built APIC. That's a conflict of interest, so here are the ground rules:

  • APIC's column comes from reading its actual source code (about 5,800 lines). I have not pixel-peeped output quality against the others.
  • Competitor facts come from their public info and 2026 review/pricing roundups. Review sites disagree on prices, so check each vendor's pricing page before you buy anything.
  • Scores are my opinion, on a 1 to 10 scale. APIC does not win overall. I'd rather lose honestly than win a rigged fight. ๐Ÿ˜„

๐Ÿงฌ Tale of the Tape

๐ŸŸข APIC Web ๐Ÿ”ต Photopea ๐ŸŸ  Pixlr ๐ŸŸฃ Canva
Nickname The Private Darkroom The PSD Monster The AI Credit Machine The Design Giant
Born for Private edits, colour, cut-outs, compression Photoshop-style layered editing Fast edits plus AI generation Social posts, decks, brand design
Runs where Browser + Windows app Browser Browser + apps Browser + apps
Your photo goes to... Stays on your device Reported on-device Cloud (AI features) Cloud
Account to edit? No No Optional Yes
Account to export? Google sign-in (one tap) No Optional Yes
Price Free, no paid tier Free with ads, Premium about $5/mo* Free tier, paid from about $1.49/mo (annual)* Free, Pro about $15/mo*

*As reported by review sites in 2026. Verify before quoting.

๐Ÿ”” ROUND 1: Price & Access

What matters: can you just open it and work?

  • Photopea wins the access battle: no account, free, and you pay in ads instead of dollars.
  • APIC has no ads, no paid tiers, no credits. Every tool is open to guests, and Google sign-in (one tap) is needed only at export.
  • Pixlr gives a free tier but meters AI with credits (around 20 free ones) and pushes upgrades.
  • Canva has a generous free plan, but its Background Remover and Magic Eraser sit behind Pro. 
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘ 8
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Photopea (no sign-in beats one-tap sign-in)

๐Ÿ”” ROUND 2: Privacy

This is the round APIC was built for.

  • APIC: the code shows images processed entirely in the browser: the filters, the AI background removal and the object removal all run on-device. The AI models download once and are cached. The honest footnote: APIC does keep your Google profile and some visit analytics.
  • Photopea: reported to process on-device too.
  • Pixlr / Canva: cloud-based. Fine for memes, less fine for ID scans, client work or anything under NDA. 
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘ 8
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 4
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: APIC

๐Ÿ”” ROUND 3: Layers & Pro Editing Depth

Here the Photoshop-style tools come out.

  • Photopea is the heavyweight: full layers, masks, selections, healing and clone tools, all with a Photoshop-like workflow.
  • Pixlr (the Pixlr E editor) is a solid middle ground.
  • APIC has a real layers panel: image and text layers with move/scale/rotate, blend modes, opacity, per-layer effects and 60 history steps. It does not have layer masks, lasso/magic-wand selection or clone/healing tools yet.
  • Canva layers things like a design tool, not a pixel editor. 
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Photopea (it isn't close, and I'm not going to pretend)

๐Ÿ”” ROUND 4: Colour Science

APIC's secret weapon is a photographer's toolkit that's free and on-device:

  • Live histogram
  • RGB curves
  • White balance (temperature + tint)
  • 8-band HSL mixer
  • 3-way colour grading
  • Clarity, vibrance, sharpen, vignette, grain, fade, gamma
  • Export your grade as a .cube LUT (33ร—33ร—33) to use in video editors

I couldn't find .cube export in the other three. Photopea answers with a deep, Photoshop-class adjustment menu, so this one is a draw at the top. Pixlr offers the basics, and Canva offers sliders and presets.

APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Canva     โ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 3
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: DRAW, APIC and Photopea

๐Ÿ”” ROUND 5: Free AI Cut-outs & Eraser

This round is about free AI cut-outs with no limits, not generative magic.

  • APIC: AI background removal (IS-Net) in three quality tiers (about 5 MB / 44 MB / 176 MB models) and an AI object remover (MI-GAN about 28 MB, or LaMa about 208 MB for harder jobs). No credits, no daily cap, no upload.
  • Photopea: one-click AI tools exist.
  • Pixlr: powerful, but AI runs on credits.
  • Canva: the best-known tools, but Pro-only.

The tradeoff: APIC's first run downloads a model (5 to 208 MB). After that it's cached.

APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 4
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: APIC (on access and limits, not a quality benchmark)

๐ŸŽค Special guest: remove.bg

A one-trick specialist that is cloud-based by design (you upload your photo). One 2026 roundup lists a small free allowance, around 3 images a month. Great at its single job, but it's a tool, not an editor.

๐Ÿ”” ROUND 6: Generative AI Power

Time for text-to-image, generative fill, AI everything.

  • Pixlr reports 15+ AI models across image, video and audio.
  • Canva leans hard into AI with Magic Studio and its newer Canva AI release.
  • Photopea has AI tools including content replacement.
  • APIC has no generative AI. It does cut-outs and erasing, not image creation. 
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
APIC      โ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 2
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: DRAW, Pixlr and Canva ๐Ÿฅถ

๐Ÿ”” ROUND 7: File Format Support

  • Photopea opens PSD, AI, RAW and more, which makes it the "open anyone's file" champion.
  • Pixlr / Canva: broad support for everyday formats and their own project files.
  • APIC exports PNG, JPEG, WebP, BMP, ICO and PDF. No PSD, RAW or HEIC import that I could find. 
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Photopea

๐Ÿ”” ROUND 8: The Utility Belt (Resize โ€ข Convert โ€ข Compress)

Developers live in this round. It's the boring task you do 20 times a week:

  • APIC: resize by px or %, convert between formats, and compress to a target file size in KB. Hit "under 100 KB" and it searches for the right quality for you.
  • Photopea / Pixlr / Canva: possible, but not the point of those tools.

๐ŸŽค Special guest: Squoosh

Google's open-source Squoosh runs its codecs in your browser and is a compression specialist, with modern formats like WebP and AVIF and a side-by-side before/after slider. In its lane it's superb. It just doesn't edit: no layers, no colour tools, no cut-outs.

APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘ 9
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 5
(Squoosh would score 10 in compression, and zero in everything else.)
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: APIC (all-rounder) / Squoosh (pure compression)

๐Ÿ”” ROUND 9: Speed to First Edit & Learning Curve

  • Canva: drag, drop, template. The easiest on earth.
  • APIC: drop an image and every tool is right there. No account wall.
  • Pixlr: friendly and quick.
  • Photopea: powerful, but it's a Photoshop clone, so expect a learning curve. 
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘ 8
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Canva

๐Ÿ”” ROUND 10: Mobile & Platforms

  • Canva: best-in-class mobile apps.
  • Pixlr: good mobile support.
  • APIC: works in a mobile browser, with a size cap on phones (about 12.6 MP) to keep things from crashing. Desktop has no cap, and there's a Windows app.
  • Photopea: powerful, but cramped on a phone. 
Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘ 8
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘ 7
Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 4
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Canva

๐Ÿ”” ROUND 11: Templates & Design Ecosystem

Not a fair fight. Canva has millions of templates, stock assets, brand kits and real-time collaboration. Pixlr has a template library behind its paid plans. Photopea and APIC are editors, not design suites.

Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘ 6
Photopea  โ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 2
APIC      โ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 1
Enter fullscreen mode Exit fullscreen mode

๐Ÿ† Round winner: Canva (flawless victory)

๐Ÿ“Š FINAL SCOREBOARD

Round ๐ŸŸข APIC ๐Ÿ”ต Photopea ๐ŸŸ  Pixlr ๐ŸŸฃ Canva
1. Price & access 8 9 6 5
2. Privacy 9 8 5 4
3. Layers & pro editing 5 10 7 5
4. Colour science 9 9 5 3
5. Free AI cut-outs 9 7 6 4
6. Generative AI 2 6 9 9
7. Formats 5 10 7 7
8. Utility belt 9 6 5 5
9. Ease & speed 8 6 7 10
10. Mobile 7 4 8 10
11. Templates 1 2 6 10
TOTAL (/110) 72 77 71 72 
๐Ÿฅ‡ Photopea  โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘ 77
๐Ÿฅˆ APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 72   (tied)
๐Ÿฅˆ Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 72   (tied)
4๏ธโƒฃ Pixlr     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘โ–‘โ–‘โ–‘โ–‘โ–‘ 71
Enter fullscreen mode Exit fullscreen mode

Photopea takes the overall belt. It's the most capable free editor on the web, and that's no accident.
But look at where APIC wins: privacy, free AI cut-outs, colour science (tie) and the utility belt. Those are the exact things a lot of developers and creators do every single day.

๐ŸŽฏ Choose Your Fighter

If you... Pick
Edit PSD files or need layers, masks and heavy compositing ๐Ÿ”ต Photopea
Need generative AI and fast creative edits ๐ŸŸ  Pixlr
Make posts, decks, thumbnails and want templates ๐ŸŸฃ Canva
Care that your photos never leave your device, want free unlimited AI cut-outs, a pro colour toolkit or compress to an exact KB ๐ŸŸข APIC
Only need to shrink images for the web ๐ŸŽค Squoosh
Only need one background removed and don't mind uploading ๐ŸŽค remove.bg

๐Ÿง  What I learned building APIC against these giants

  1. You can't out-feature Photopea. APIC competes by being the private, no-credits, no-ads option, not by cloning Photoshop.
  2. "Free" means very different things. Ads, credits, paywalled AI and mandatory accounts are all "free". APIC's price is one Google sign-in at export. That's a deliberate trade, and it keeps the tool free.
  3. On-device AI is real now. Background removal and object removal running in a browser tab, with no server, would have sounded impossible a few years ago.
  4. The roadmap is obvious: PSD and HEIC import, layer masks, selection tools and batch processing.

๐Ÿ”— Try the underdog

Your turn: which editor do you actually use, and which round did I get wrong? Tell me in the comments and I'll update the scoreboard. ๐ŸฅŠ

Disclosure: I'm the creator of APIC. Scores are subjective, competitor details come from public information and may have changed, so verify pricing and limits before you decide.

APIC
---An Akhouri Systems product

Top comments (49)

Collapse
 
effessdev profile image
EffessDev •

Bro why this?

Canva     โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ 10
APIC      โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–‘โ–‘ 8
Enter fullscreen mode Exit fullscreen mode

Isn't APIC a bit easier?

Collapse
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bro but when I pasted my code in all AIs(literally All then they all said canva is more easier I don't know) I also thought that APIC is more easy but still I felt like AIs have data which My mind don't have in billions ๐Ÿ˜‚

Collapse
 
effessdev profile image
EffessDev •

Yeah I also think APIC is easier...

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

hmm so APIC won against photopea overall?

Thread Thread
 
effessdev profile image
EffessDev •

In certain cases only ๐Ÿ˜… But in certain cases it does (because Photopea freeze my PC at times APIC doesn't do that now though it used to do previously but you fixed it!)

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

yeah.

Thread Thread
 
effessdev profile image
EffessDev •

So shall we discuss the rest of your workflow here? ๐Ÿ˜…

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

hmm okay but I don't want to leak my tactics publicly

Thread Thread
 
effessdev profile image
EffessDev •

Ok no... ๐Ÿคฃ

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

Thread Thread
 
effessdev profile image
EffessDev •

So, you said that ATLOCK is a single 10k LOC python file, right?

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

yepp

Thread Thread
 
effessdev profile image
EffessDev •

But why a single file? Why don't you split it into multiple files? I can't even handle this 2.5k LOC file.

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bro, It is intentionaly deseigned as a single file architecture to simplify compiling into a standalone .exe and minimize the attack surface for local file-tampering....and there are some more minor reasons....
last time when I tried modulation for APIC desktop ones I got cooked but I think in v6 I will keep it module bise because I feel like I'm not that old AAK who built APIC v1 that time.

Thread Thread
 
effessdev profile image
EffessDev •

Bro you know how to make dynamic island using Qt yourself but don't know how to do imports? I don't believe that... 10k LOC is more than what can be accepted for a single file.

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

lol I didn't said I can't, I can bro but I just preffered it also
It's not justt about knowing imports broo,it's a importantt security choice: a single file means one attack surface, no import-chain hijacking, and trivial self-integrity verification...,
Splitting it into modules would eventually lead to an attacker swap individual files withoutt breaking the App's overall hash, which completely defeats the point of a tamper..evident security suite.
that's the point behind it if knowledge was the point then Akhouri Systems never was being founded. ๐Ÿคฃ๐Ÿคฃ

Thread Thread
 
effessdev profile image
EffessDev •

Yeah ok whatever... Still not convinced. ๐Ÿค” I'll be back soon (coffee break)...

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

come fast broooo

Thread Thread
 
effessdev profile image
EffessDev •

๐Ÿ‘

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bro you became 2nd reader of that 31august post, 1 was me myself lol.
in that time no one read it.

Thread Thread
 
effessdev profile image
EffessDev •

Unfortunately we didn't meet at that time otherwise I'd have read it :(

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

yeah, but we are now BROSS

Thread Thread
 
effessdev profile image
EffessDev •

yeah ๐Ÿ˜…

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bro you are online then why are not discussing anything...

Thread Thread
 
effessdev profile image
EffessDev •

I was waiting for you to say something ๐Ÿ˜ข

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bruhh, and I was waiting for you and always refreshing dev.to waiting for notification

Thread Thread
 
effessdev profile image
EffessDev •

Great! ๐Ÿคฃ

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

@effessdev again waiting for me to say something?

Thread Thread
 
effessdev profile image
EffessDev •

I had dev.to/notifications tab open. But unfortunately I have to refresh it every time.

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

hmm ;

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

"exams are many, my marks are none ๐Ÿ˜‚"

Thread Thread
 
effessdev profile image
EffessDev •

What? Didn't you get 3rd rank in your class?

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

yeah got 3rd rank but still a meme lol.
and the 3rd rank I got is purely wasted I can clearly see in my home ๐Ÿ˜‚
literally they just said that's not big thing maintaining it is big thing also they said more stuffs(stupidity on peak ๐Ÿ–๏ธ๐Ÿคš๐Ÿ”ฅ)

Thread Thread
 
effessdev profile image
EffessDev •

Bro this is public they can read all of this ๐Ÿค

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

lol I don't think IT head and ex IT head both have account in dev.to ๐Ÿ˜‚

Thread Thread
 
effessdev profile image
EffessDev •

What if they do? ๐Ÿ‘๏ธโ€๐Ÿ—จ๏ธ

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

they at max to max they can delete me from house ๐Ÿซก
at least then I'll have peace

Thread Thread
 
effessdev profile image
EffessDev •

BROOO you really don't have boundaries ๐Ÿคฏ in everything

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

which boundary?

Thread Thread
 
effessdev profile image
EffessDev •

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

Thread Thread
 
effessdev profile image
EffessDev •

Why didn't you put your dream phone? ๐Ÿคฃ Is this your dream phone now?

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

Collapse
 
merbayerp profile image
Mustafa ERBAY • • Edited

BROOOOO, Anmol! ๐Ÿ˜‚๐Ÿ”ฅ

First ATLOCK v5, and now APIC Web?

Youโ€™re keeping me busy, little brother! At this rate, Iโ€™ll need to open a dedicated testing department just for Akhouri Systems. ๐Ÿ˜‚โค๏ธ

Jokes aside, I read your comparison carefully, and thereโ€™s something I genuinely appreciate: you openly acknowledged your conflict of interest, explained your scoring methodology, and didnโ€™t automatically declare your own product the winner.

Thatโ€™s a good starting point for an engineering discussion.

However, you asked which rounds you got wrong and invited people to challenge your conclusions.

Well, brother, I decided to go a little further. ๐Ÿ˜„

Rather than simply comparing feature lists, I examined APIC Webโ€™s implementation and reviewed its behavior against several of its advertised capabilities.

Independent review date: October 21, 2026

Reviewed commit: 02eabd45417fe5c6734807e07f75b507fbb80122

Review scope: Live browser testing, source inspection, isolated algorithm testing, functional correctness, and privacy architecture.

All test images were synthetic. No personal photographs or Google accounts were used.

Here are my findings.

โธป

๐Ÿ”ด Finding 1 โ€” Cropping Reapplies Existing Colour Adjustments

Severity: P2 โ€” Functional correctness

Status: Confirmed in the live browser and supported by source analysis.

This is the most interesting functional issue I found.

I created a synthetic 400 ร— 300 PNG containing a uniform RGB(128,128,128) grey background.

Reproduction steps:

  1. Open the image in APIC as a guest.
  2. Navigate to Edit.
  3. Set Brightness to 50%.
  4. Switch to Advanced.
  5. Select a rectangular crop.
  6. Apply the crop.
  7. Observe the resulting image.

Expected behavior:

Cropping should change the image boundaries without unexpectedly changing the appearance of the remaining pixels.

Actual behavior:

The cropped image becomes noticeably darker, even though Brightness still displays 50%.

The resulting dimensions changed from 400 ร— 300 to 220 ร— 150.

Measured from the rendered screenshots:

  • Before cropping: RGB(56,56,56)
  • After cropping: RGB(29,29,29)

These are screenshot/display measurements, not direct canvas pixel measurements.

The idealized filter calculation illustrates the issue:

128 ร— 0.5 = 64

64 ร— 0.5 = 32

In other words, the brightness adjustment is effectively applied twice.

Root cause:

In index.html, lines 2252โ€“2265, the crop operation copies pixels from the already-rendered canvas into a new image and assigns that result to originalImg.

It then calls render() without resetting the active adjustment state.

The rendering pipeline subsequently reapplies currentFilterString() and applyAdvancedFx().

Interestingly, the resize workflow uses bakeToOriginal(), which resets the relevant adjustment state.

The crop workflow does not follow the same approach.

Suggested fix:

Choose one consistent editing model:

  • Crop the unadjusted source and preserve the adjustment state.
  • Or bake the visible result into the source and reset adjustments that have already been applied.

Whichever approach you choose, keep undo snapshots consistent.

I would also recommend regression tests covering brightness, contrast, curves, and white balance.

Source references:

index.html:2252โ€“2265

index.html:1995โ€“2042

This is a reproducible functional bug, not a security vulnerability.

โธป

๐ŸŸ  Finding 2 โ€” Target-Size Compression Can Silently Exceed Its Limit

Severity: P2 โ€” Functional correctness

Status: Confirmed in an isolated algorithm test; actual signed-in browser export not tested.

Your compression feature is one of APICโ€™s most useful capabilities.

The ability to request an output under a specific file size is particularly valuable for developers preparing website assets.

However, the current algorithm has an important edge case.

The function findQualityForTarget() searches for an acceptable compression quality between approximately 0.05 and 0.98.

When the requested target cannot be achieved, the function falls back to quality 0.05.

The problem is that it returns this result without verifying that the final encoded image actually satisfies the requested size.

I tested the original search logic using a deterministic 1024 ร— 768 high-entropy RGB image.

For this isolated test, the browserโ€™s canvasToBlob() implementation was replaced with a Pillow JPEG encoder.

Results:

Test A โ€” Reachable target

Requested: 102,400 bytes

Returned: 100,278 bytes

Result: Within target.

Test B โ€” Unreachable target

Requested: 1,024 bytes

Returned: 51,165 bytes

Result: Approximately 50 times the requested size.

The function returned normally without reporting that the target was impossible.

To be precise, these measurements came from the isolated algorithm test, not a completed browser download. The browser encoder may produce different byte counts.

Nevertheless, the control-flow problem is clear.

Source references:

index.html:3917โ€“3926

index.html:4090โ€“4098

Suggested fix:

After compression, explicitly compare the encoded blob size against the requested limit.

If the target cannot be reached:

  1. Inform the user that the requested size is unattainable at the current dimensions.
  2. Offer an optional resolution reduction.
  3. Display the actual resulting file size.
  4. Avoid silently presenting an oversized output as a successful target-size export.

A useful regression test would attempt to compress high-entropy images to deliberately impossible targets.

One additional point: the signed-in export flow still needs a browser-level reproduction test before this finding can be described as an observed downloaded-file failure.

โธป

๐ŸŸก Finding 3 โ€” History Capacity Is Described Inaccurately

Category: Documentation accuracy

Your comparison mentions 60 history steps.

However, the implementation distinguishes between the main photo history and the layer history.

The main photo history is capped at:

  • Desktop: 40 snapshots
  • Normal mobile: 10 snapshots
  • Low-tier mobile: 5 snapshots

The layer history has a separate 60-snapshot limit.

Also, the initial state occupies a snapshot, so the number of available undo operations may be lower than the snapshot count.

Source references:

index.html:2047

index.html:5364

I would recommend describing these separately.

Something like:

โ€œUp to 40 main-image history snapshots on desktop, with device-dependent mobile limits and a separate 60-snapshot layer history.โ€

This isnโ€™t a vulnerability, but accurate documentation matters when comparing products.

โธป

๐ŸŸก Finding 4 โ€” Mobile Image Limits Need More Precise Documentation

Category: Documentation accuracy

Your article mentions an approximately 12.6-megapixel mobile limit.

The implementation also includes a long-side dimension restriction through mobileSideCap().

This means an image can potentially be downscaled even when its total pixel count is below the advertised megapixel threshold.

Source reference:

index.html:1912 onward

I would suggest documenting both restrictions.

For example:

โ€œMobile images are subject to pixel-count and maximum-dimension limits, depending on device capability.โ€

No physical mobile-device testing was performed during this review, so this observation comes from source inspection.

โธป

๐Ÿ” Finding 5 โ€” Privacy Architecture: Promising, but Claims Need Precise Boundaries

Status: Source-reviewed; no demonstrated photo exfiltration.

This is where APIC genuinely deserves credit.

The inspected implementation uses browser-local image processing through canvas, blobs, and local AI inference.

The AI models are downloaded for execution on the client side.

I did not identify a photo-upload request in the inspected application-level processing paths.

That is consistent with your on-device image-processing claim.

However, thereโ€™s an important distinction between:

โ€œPhotos are processed locally.โ€

and

โ€œThe application makes no network requests.โ€

These are not equivalent.

APIC also integrates:

  • Google Identity
  • Firebase Authentication
  • Firestore
  • Google Fonts
  • External CDN and model hosts
  • Visit tracking and counters

The inspected startup logic invokes trackVisit() and loadCounters().

Signed-in profile handling includes information such as name, email address, profile-photo URL, and sign-in metadata.

To your credit, your article already acknowledges profile data and analytics.

Iโ€™m not presenting those integrations as vulnerabilities.

Also, public Firebase client configuration is not automatically a leaked server credential.

My recommendation is simply to make the privacy guarantees more explicit.

For example:

โ€œImage processing runs locally in the browser. Authentication, analytics, model downloads, and supporting services may communicate with external endpoints.โ€

That is a stronger and more technically defensible privacy statement.

Important limitation:

I did not capture a complete network request/body trace or independently audit every third-party SDK.

Therefore, this review does not prove that image bytes can never leave the browser under every possible execution path.

No image exfiltration was demonstrated.

โธป

โœ… What Worked Correctly

Now for the good news, brother! ๐Ÿ˜„

I donโ€™t believe in publishing only failures.

A fair engineering review should also document what actually works.

The following checks passed during the live browser assessment:

  1. Image import

A synthetic 800 ร— 600 PNG loaded successfully with its expected dimensions.

  1. Rotation

A 90-degree rotation correctly changed the canvas from 800 ร— 600 to 600 ร— 800.

  1. Undo

Undoing the rotation restored the original dimensions.

  1. Aspect-ratio-preserving resize

Setting the width to 400 correctly produced a height of 300.

Applying the resize produced the expected 400 ร— 300 result.

  1. Layer operations

Text-layer creation and duplication worked.

Undo successfully removed the duplicated layer.

Detailed text-edit history was not fully validated, so Iโ€™m not making broader claims about that behavior.

  1. Guest export access control

Clicking Export image as a guest displayed the sign-in requirement and the option to decline.

I did not attempt to bypass authentication.

  1. AI background removal

The Light background-removal mode completed on a synthetic geometric image.

The application reached its completion state and retained the expected 800 ร— 600 canvas dimensions.

No warning or error was captured during that smoke test.

However, this was not a comprehensive AI quality benchmark.

I did not independently verify the resulting alpha channel, offline caching, or all three model variants.

Still, the basic workflow completed successfully.

And thatโ€™s worth acknowledging. ๐Ÿ‘

โธป

๐Ÿงช What Remains Untested

For transparency, the following areas remain outside this reviewโ€™s verified scope:

  • Signed-in PNG, JPEG, WebP, BMP, ICO, and PDF exports
  • Actual downloaded-file integrity
  • LUT export correctness
  • All AI background-removal quality modes
  • AI object removal
  • Offline behavior and model caching
  • Physical mobile-device performance
  • Malformed or adversarial image inputs
  • Backend authorization rules
  • Complete third-party dependency auditing
  • Controlled image-quality comparisons against competitors

I also did not independently reproduce your Photopea, Pixlr, or Canva benchmarks.

Therefore, Iโ€™m not assigning new comparative scores based on these tests.

โธป

๐Ÿ“Š My Overall Assessment

Anmol, I think APIC has a sensible product direction.

Trying to outperform Photopea at every feature would be an enormous undertaking.

But offering a focused, accessible, browser-based editor with local AI processing, useful colour controls, and practical compression tools is a reasonable strategy.

The important next step is improving correctness and reliability.

My recommended priorities would be:

P2 โ€” Fix the crop adjustment pipeline.

Ensure geometric operations donโ€™t unexpectedly apply existing colour adjustments multiple times.

P2 โ€” Enforce compression target-size guarantees.

An impossible target should produce an explicit outcome rather than silently returning an oversized result.

Documentation โ€” Clarify history and mobile limits.

Keep published claims aligned with actual implementation behavior.

Privacy โ€” Document the exact trust boundaries.

Clearly distinguish local image processing from authentication, analytics, and external resource requests.

Quality assurance โ€” Expand automated regression coverage.

Especially for operation ordering, image-state consistency, undo/redo, compression edge cases, and export integrity.

โธป

โค๏ธ One Final Thought

Brother, after more than 20 years working with systems architecture, enterprise infrastructure, and software engineering, Iโ€™ve learned that building features is only one part of building a reliable product.

The difficult part is making those features behave consistently when users combine operations in unexpected ways.

Crop after brightness.

Resize after rotation.

Undo after changing layers.

Compress an image to an impossible target.

These ordinary workflows are where subtle implementation mistakes become visible.

And thatโ€™s exactly why independent testing matters.

Review summary:

  • 1 functional bug reproduced in the live browser.
  • 1 compression algorithm defect confirmed through isolated testing.
  • 2 documentation discrepancies identified.
  • Multiple core workflows passed their basic functional checks.
  • Local image-processing claims are consistent with the inspected code paths.
  • No confirmed security vulnerability or photo exfiltration.
  • Several advanced workflows remain untested.

I hope these findings help make APIC stronger.

And just like with ATLOCK, Iโ€™m not interested in your โ‚น50, brother. ๐Ÿ˜‚โค๏ธ

Keep your money and invest it in the next version.

What matters to me is encouraging good engineering, sharing reproducible evidence, and helping developers improve their work.

You invited people to challenge your product, and I respect that.

Now you have a few concrete test cases for your roadmap. ๐Ÿ˜„

One last thing:

Donโ€™t build APIC to win a subjective scoreboard. Build it so that its strongest claims survive independent testing.

Thatโ€™s a much more valuable victory.

Keep building, keep learning, and keep questioning your assumptions.

Much respect from Tรผrkiye, little brother! ๐Ÿ‡น๐Ÿ‡ท๐Ÿค๐Ÿ‡ฎ๐Ÿ‡ณ

And please donโ€™t release APIC and ATLOCK v6 on the same day.

My testing queue needs a vacation! ๐Ÿ˜‚๐Ÿ”ฅ

Collapse
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

bro who are youuuuuuuuuuuuuuuuuuuuuuu
you break ATLOCK v5 like a joke few minutes ago, now APIC Web
bro honestly saying, you deserve to flex your 20year+ stuff bro you are god I've no words brooo this will help me a lot to make APIC Web improved more...
bro thank you so muchhhhh
I know maybe no one accept my challenge and help me to improve my products but I also know there is one man army standing behind akhouri systems always that's you broo
Thanks multiplied by googol times broooo

Collapse
 
merbayerp profile image
Mustafa ERBAY •

Mustafa Erbay. ๐Ÿ˜Ž

Thread Thread
 
akhourianmolkumar profile image
Akhouri Anmol Kumar •

in hindi: เคจเคพเคฎ เค•เคพ เค…เคชเคจเคพ เคเค• เค…เคฒเค— เค“เคœ(aura) เคนเฅˆเฅค
in english: the name itself carries an aura

Thread Thread
 
merbayerp profile image
Mustafa ERBAY •

Haha, youโ€™re giving me too much credit, brother! ๐Ÿ˜‚โค๏ธ Iโ€™m just an engineer who still loves learning and sharing knowledge. Experience means little if we donโ€™t use it to help each other grow. Keep building, my friend! ๐Ÿค๐Ÿ”ฅ

View full discussion (49 comments)