What the measured run showed
These figures describe the controlled fixture documented for this guide. They are evidence for the tested case, not a universal performance guarantee.
A case where PNG was already smaller
Our 1280 × 800 synthetic format fixture started as a 31,794-byte PNG. Converting that exact image through the tested NetroDoc paths produced a 370,186-byte JPG and a 200,460-byte WebP. In other words, both converted files were larger than the source PNG in this case.
That does not mean PNG is generally smaller than JPG or WebP. This fixture contained large, regular regions and patterns that PNG compression handled unusually well. The result is useful precisely because it shows why format choice should follow image content rather than a blanket rule.
- PNG source: 31,794 bytes.
- JPG output: 370,186 bytes, PSNR 23.70 dB against the source fixture.
- WebP output: 200,460 bytes, PSNR 23.67 dB against the source fixture.
- Dimensions remained 1280 × 800 in both conversions.
Transparency changes the decision
In a separate transparency fixture, the source had an alpha channel. WebP retained alpha information, while JPG became RGB and the transparent corner was flattened to white in the tested conversion.
That makes transparency a functional requirement, not a cosmetic preference. Logos, overlays, interface assets, and cut-out graphics can lose intended appearance when forced into JPG.
- Use PNG when lossless pixels or predictable alpha support are important.
- Use WebP when you need alpha support but also want a modern compressed format.
- Use JPG for photographic content when transparency is not required and some lossy compression is acceptable.
What our JPG compression fixture measured
The JPG compression fixture started at 966,381 bytes. With the tested quality setting of 75, the output was 214,727 bytes: a 77.78% reduction for that specific image. The measured PSNR against the input was 22.76 dB.
A size reduction is not free. JPG compression is lossy, so repeated saving can accumulate artifacts. For screenshots, text-heavy graphics, diagrams, or images that will be edited again, the smallest file is not always the best result.
What our PNG compression fixture measured
The PNG compression fixture fell from 5,042,630 bytes to 7,199 bytes, a 99.86% reduction, while the decoded pixels remained identical in our comparison.
That percentage is deliberately not presented as a normal expectation. The fixture was synthetic and extremely compressible. The trustworthy conclusion is narrower: the tested PNG path can reduce storage overhead without changing pixels when the source encoding is inefficient, but savings depend strongly on the source image.
A better format decision rule
Start with the visual requirement, then optimize size. If you need exact pixels or transparency, PNG is a strong default. If the content is photographic and a little loss is acceptable, JPG is often practical. If you need modern lossy or lossless-style compression with alpha support, WebP is worth testing against the actual source rather than assuming it will win.
When file size matters, compare the converted output against the original at the dimensions you actually plan to publish. A format conversion that increases bytes without delivering another benefit is not an optimization.
Temporary processing and privacy
Image conversions use temporary request working directories that are removed after the response lifecycle completes. The separate 3,600-second TTL applies to queued conversion and OCR jobs, not ordinary image conversion requests.
Read the Privacy Policy →Troubleshooting
WebP or JPG became larger than PNG
The source may already be efficiently compressed or contain patterns that PNG handles well. Compare actual output size instead of assuming the target format will be smaller.
Transparency disappeared after JPG conversion
JPG has no alpha channel. Use PNG or WebP when transparency is part of the image.
JPG looks worse after repeated edits
Lossy re-encoding can accumulate artifacts. Keep a high-quality or lossless master and export compressed copies from that master.
Method note
NetroDoc publishes measured claims only after a controlled run completes successfully. Untested cases are kept out of the conclusion. The evidence run for this publication wave used the same production runtime components while operating on an isolated candidate and temporary server paths.