Maria Rodriguez
DevOps Engineer · Performance Specialist
DevOps Engineer · Performance Specialist
I'm Maria. I've spent the last 8 years optimizing web infrastructure — CDN configs, image pipelines, build tooling — for companies where a 500ms slowdown directly hits revenue. Compress2PNG is the tool I built after getting frustrated that most "image compressors" give you a percentage reduction but no way to verify the actual impact on page speed.
Two things. First, everything runs in your browser — zero uploads, zero server processing. You can verify this by going offline mid-session. Second, the before/after comparison slider isn't just cosmetic. It shows you exactly what data you're losing at each quality level, so you can make an informed trade-off instead of guessing at a quality slider.
I track these metrics for every image I compress:
| Metric | Why It Matters |
|---|---|
| File size reduction % | Bandwidth savings — directly impacts hosting costs and load time |
| SSIM (structural similarity) | How much visual quality is actually lost — the number that matters, not just "80% quality" |
| LCP impact | If this image is above the fold, compression directly improves Google's Core Web Vitals score |
| Encode time | Important for batch workflows — some formats are 10x slower than others at similar quality |
Most compressor tools only show you #1. That's like telling you a diet plan reduced your weight without mentioning you lost all your muscle.
I run this on Cloudflare Pages — the cost is negligible at this scale. The site is ad-supported because that keeps the incentives simple: make a useful tool, get traffic, serve ads. No tracking scripts, no email collection, no "growth hacks." I'm not building a startup. I'm hosting a tool.
Before this, I worked on image delivery pipelines at a mid-size ecommerce platform. I've seen firsthand that switching from JPEG to WebP cuts ecommerce image bandwidth by 30-40% — and that dropping Largest Contentful Paint from 4s to 2s increases conversion rates by 8-12%. The articles on this blog are informed by those production experiences, not by scraping Google results and running them through an AI summarizer.
Found a bug? Have compression benchmarks to share? [email protected]