Convert PNG to WebP — Lossless for UI, Lossy for Photos, Decided Per File

The one-pass fix for a mixed PNG folder. Screenshots go lossless. Photos go lossy-with-alpha. One drop zone, one batch, one ZIP out — nothing uploaded.

2 paths chosen per image
no slider quality picked automatically
0 files uploaded, ever
Start converting

A router, not a slider

Drop your PNGs — mixed or otherwise. Every file gets the path that makes it smaller at the quality bar. You don't pick. You don't sort. You get a ZIP back.

Supported input formats

  • JPG / JPEG — Photos, portraits, web content
  • PNG — Screenshots, icons, transparent images
  • HEIC / HEIF — iPhone photos, Apple formats
  • TIFF — Scans, prints, high-resolution archives
  • GIF — Animations and static GIFs
  • BMP, PSD & more — Anything ImageMagick can decode

How the conversion works

  1. 1. Drop
    Drag files or a whole folder into the box below. Folder structure is preserved in the output ZIP.
  2. 2. Analyze
    Each image is analyzed for entropy and content type. The engine picks per-image quality settings targeting PSNR ≥ 44.5 and SSIM ≥ 0.95.
  3. 3. Encode
    Conversion runs on all of your CPU cores in parallel via Web Workers. EXIF, ICC color profiles and geolocation are copied onto the WebP or AVIF output.
  4. 4. Download
    When the batch is done, a ZIP containing every converted file downloads automatically. No re-upload, no waiting on a server.

Why WebP is the right target for a PNG archive

PNG is two formats sharing one file extension: a lossless codec for synthetic graphics, and a last-resort container that people accidentally export photos into. WebP has a clean answer for each case, so you don't have to sort.

PNG is two formats sharing one extension

Open any real PNG folder and you'll find two different things cohabiting. One pile is synthetic — screenshots, UI mockups, logos, diagrams, flat icons — where PNG was genuinely doing useful work. The other pile is photographs someone exported as PNG by mistake, sitting there at 15 MB each, taunting your disk. Most tools ask you to sort them. This one doesn't.

A router, not a slider

Each file gets inspected. If it's flat and synthetic, it goes through the lossless path and lands 20–30% under the source with every pixel intact. If it's photographic, it takes the lossy-with-alpha route and drops 70–90% of the weight without a visible artefact. The decision is per-file and automatic. You don't pre-sort. You don't pick an encoder. You drop the folder.

Alpha is a first-class channel

Both paths carry the alpha plane end-to-end. The feathered edge on a cut-out logo, the gradient under a drop shadow, the 40% overlay in a product mockup — they survive. If the source turns out to be fully opaque (a surprising number of 'transparent' PNGs are), the alpha plane gets dropped and the output shrinks a little more.

What survives the trip

Your Photoshop export with a Display P3 profile still looks right on a calibrated monitor. ICC, EXIF and XMP are lifted during decode and rewritten onto the WebP. DPI hints carry across where WebP has a matching field; anything it can't represent gets dropped honestly rather than faked.

What actually changes when you re-encode

Criterion PNG WebP
UI screenshot / flat graphic baseline ~20–30% smaller (lossless)
Photo-ish PNG (24-bit gradients) baseline 70–90% smaller (lossy + alpha)
Transparent pixel handling 8-bit alpha 8-bit alpha, full fidelity
Encoder path chosen DEFLATE, always Lossless or lossy-with-alpha, per file
Metadata (ICC / EXIF / XMP) stored in PNG chunks copied to output
Browser support universal universal (Safari 14+, all others)
Animated input (APNG) rare native support first frame only

How SciZone converts PNG to WebP

Drop the folder, let the per-image router pick the path, download the ZIP.

  1. 1

    Drop the whole archive

    Drag a folder of PNGs — mixed screenshots, logos, Figma exports, stray photo-PNGs — onto the drop zone. Nested folders come out of the other side intact. A 48 MP surprise hidden in the pile won't blow the heap on arrival.

  2. 2

    Per-image routing runs

    Each file is inspected — entropy, resolution, whether the alpha channel is doing anything non-trivial. The encoder picks the path that'll produce the smaller file at the quality bar. No slider. No dropdown. No decision fatigue.

  3. 3

    Quality picked per image, where it applies

    Photos get an adaptive quality search — a short look for the setting that's transparent to the eye on that specific image. Lossless paths skip the search entirely and just write the compressed bitstream.

  4. 4

    Grab the ZIP

    Your WebPs come out as a ZIP mirroring the input folder structure. The results page shows the lossless-vs-lossy decision and per-file byte savings so you can spot-check anything that looks surprising.

WebP

WebP Results

See how WebP compresses real photos with our Excellent preset — the default when you click convert.

Portrait — after Portrait — before
Before
After
Portrait
3000×4500
1.82 MB 0.26 MB
-86%
Foliage — after Foliage — before
Before
After
Foliage
3000×4500
3.10 MB 1.60 MB
-48%
Lifestyle — after Lifestyle — before
Before
After
Lifestyle
3000×2000
0.82 MB 0.44 MB
-47%

Typical WebP savings

Measured on 24 diverse photos at matched perceived quality (SSIM ≥ 0.95)

45-85%
Typical size reduction
SSIM ≥ 0.95
Perceptually matched
1000+
Files per batch

PNG to WebP — questions teams actually ask

I have screenshots and product photos saved as PNG in the same folder — does that matter?

That's exactly the intended workload. Each file is routed independently, so a screenshot can come out as lossless WebP in the same batch that turns a photo-PNG into lossy WebP with alpha. Don't pre-sort. It'll figure it out.

How is this different from dragging PNGs into Squoosh one at a time?

Squoosh asks you to pick an encoder and a quality number for every file. This tool picks both automatically using a perceptual-quality target, and runs the batch across every core you have. On a 500-screenshot archive the human-time difference is hours, not minutes.

What does 'lossless WebP' actually save versus the source PNG?

About 20–30% on typical UI captures, sometimes more on logos and icons. Both files are pixel-for-pixel identical. If lossless happens to be bigger on a particular input (rare, but possible), the engine falls back to the lossy path automatically.

Why is there a small download the first time?

That's the WebAssembly image engine — under 5 MB to download, once. After that it's cached; subsequent visits and extra batches don't re-fetch.

Does it matter if my PNG has no transparency at all?

A tiny bit, yes. If the alpha channel is uniformly 255, the encoder drops the plane entirely and the output is a plain three-channel WebP — usually a few percent smaller than carrying an unused alpha along for the ride.

Do the workers get recycled during a long batch?

Every worker is retired after 32 jobs and a fresh one takes its place. That's what keeps a 1000-file batch from quietly leaking memory — without the recycle we saw growth past a gigabyte on long runs, the kind of thing you only notice once the tab crashes.

Can I force everything to lossless mode?

Not from the UI today. The per-image router is the defining feature of the tool — overriding it would make the output worse on photo-PNGs, which defeats the purpose. If you need bitwise-identical lossless output on a photo archive, re-encoding to WebP at all is the wrong step. Keep the source.

Does anything get uploaded?

No. Every encode happens inside the tab. Open DevTools → Network during a run and you'll see the engine load once and nothing else leave. The cross-origin isolation you may notice in the console is there so the encoder can thread across your cores, not for any remote call.

Why Choose SciZone?

We're not just another optimizer. We engineered a fundamentally better solution.

Feature
SciZone (You're here)
Other Optimizers
CPU Utilization
How processing power is used
True Multi-Threading Intelligently uses all CPU cores
without overloading your system
Single-Threaded Uses only one CPU core,
wastes available power
AVIF Encode Speed
How fast AVIF actually runs in the browser
Tile-Parallel Encoding Each AVIF image is split into tiles encoded
across every core — ~6× faster than
single-tile libaom on large photos
Single-Tile Default libaom's internal threading caps around
4 threads per encode, regardless of
how many cores you have
Quality Settings
How compression is optimized
Unique Per Image Algorithm analyzes each photo
and picks optimal settings
One-Size-Fits-All Same settings for every photo,
inconsistent quality
Metadata & Color Profiles
Preservation of image data
Fully Preserved EXIF, color profiles, geolocation.
Everything stays intact
Often Stripped Color profiles lost,
metadata incomplete
Quality-Size Balance
Optimization results
Perfect Balance Maximum compression with
imperceptible quality loss
Inconsistent Either too large or
noticeable quality loss

The Bottom Line

Every photo is unique. Our intelligent algorithm understands this and analyzes each image individually to find the perfect balance between file size and quality. We utilize your computer's full power without overloading it, preserving every detail of your metadata and color profiles. Your files are smaller, faster, and absolutely perfect. 🎯