We ran real images through digimg in the browser, measured output bytes and quality (SSIM/PSNR), and captured every network request. Here's what the data actually shows — including a caveat most comparisons skip.
Method: headless Chrome driving digimg's @jsquash encoders; output quality measured with scikit-image SSIM/PSNR against the source. Higher SSIM (max 1.0) = more faithful.
| Image | Format | Size | SSIM | PSNR | Verdict |
|---|---|---|---|---|---|
| Photo (EXIF/ICC) | WebP | 19.1 KB | 0.974 | 41 dB | Excellent |
| Photo (EXIF/ICC) | JPEG | 32.5 KB | 0.980 | 38 dB | Excellent |
| Transparent logo | oxipng (lossless) | 2.7 KB | 1.000 | ∞ | Perfect, keeps alpha |
| Transparent logo | WebP | 7.1 KB | 0.998 | 47 dB | Excellent |
| Flat graphic | WebP | 7.6 KB | 0.993 | 39 dB | Excellent |
Same four source images run through all three tools. digimg measured in-browser (build 10a7517); TinyPNG (free web tier) and ShortPixel (web, Lossy/SmartCompress) measured 2026-11-15. Quality = scikit-image SSIM on RGB (max 1.0 = faithful), same metric for all. Sizes are real output bytes.
| Image (source) | digimg best | TinyPNG | ShortPixel | Takeaway |
|---|---|---|---|---|
| Photo (73.5 KB) | mozjpeg 32.5 KB · 0.980 WebP 19.1 KB · 0.974 | 52.1 KB · 0.988 | 45.5 KB · 0.978 | digimg smallest JPEG (−38% vs TinyPNG, −29% vs ShortPixel); WebP smaller still |
| Transparent logo (7.7 KB) | oxipng 2.66 KB · 1.000 | 2.63 KB · 1.000 | 2.66 KB · 1.000 | Three-way tie, all lossless (pixel-exact) |
| Flat graphic (7.3 KB) | oxipng 2.69 KB · 1.000 | 2.80 KB · 1.000 | 2.69 KB · 1.000 | digimg = ShortPixel, both lossless, beat TinyPNG |
| Screenshot UI (4.7 KB) | oxipng 1.70 KB · 1.000 | 2.82 KB · 1.000 | 1.38 KB · 0.9957 | ShortPixel smallest but lossy; digimg smallest lossless |
Photo: all three return lossy JPEGs at excellent quality (SSIM 0.978–0.988); digimg's mozjpeg is the smallest JPEG and its WebP is smaller still. PNGs: TinyPNG and ShortPixel quantize to a palette — pixel-exact for the logo/graphic (lossless); for the screenshot ShortPixel goes lossy (SSIM 0.9957) to reach 1.38 KB, while digimg oxipng stays pixel-perfect at 1.70 KB. Honest read: digimg wins on photos and lossless PNGs, ties on logos, and trades a few bytes for guaranteed lossless on screenshots — all without uploading your images.
Don't chase one compression number. WebP is the reliable high-quality choice (SSIM ≥ 0.97); oxipng is pixel-perfect for logos; PNG/JXL can grow already-compressed inputs — we show that honestly.
| Dimension | digimg | ShortPixel | TinyPNG |
|---|---|---|---|
| Where compression runs | Your browser (local WASM) | Cloud API (EU) | Cloud API |
| Image upload required | No — images stay local | Yes | Yes |
| Formats | AVIF/WebP/JXL/JPEG/PNG/oxipng | WebP/AVIF/JPEG/PNG | WebP/AVIF/JPEG XL/PNG |
| EXIF/ICC stripped | Yes (verified) | Yes | Yes |
| Price | Free | Free 100 credits/mo; Unlimited $10.99/mo* | 500 free/mo; $0.009→$0.002/compression* |
| Best for | Privacy / offline / no-account | WordPress bulk automation | Developer API pipelines |
Green = measured here. Grey = competitor published specs (*REPORTED). TinyPNG and ShortPixel same-image compression are now measured (see table above); pricing/features below remain published specs.
digimg's compression is genuinely local — image bytes never leave the browser (network-capture verified). But the digimg website loads third-party analytics and ads (Cloudflare Web Analytics, Google AdSense, and Amplitude product analytics). The accurate claim is "your images are never uploaded" — not "no third parties at all." Amplitude receives only coarse, allow-listed product events (bucketed sizes/counts, format, workflow) — never filenames, image bytes, EXIF, or data URLs. For a fully offline workflow, self-host or block analytics.
digimg — free, no-account, browser-local; best when images shouldn't be uploaded (sensitive screenshots, transparent logos via lossless oxipng, quick WebP).
ShortPixel — WordPress bulk automation where cloud upload is acceptable. TinyPNG — simple pay-per-compression developer API.
Try digimg free →digimg: real browser test (headless Chrome + CDP) on build 10a7517; output bytes from download blobs; magic-number verified; SSIM/PSNR via scikit-image; AVIF decode cross-checked in Chrome/ffmpeg/PIL; zero-image-upload confirmed via full network capture. TinyPNG: same four source images compressed via the free web tier (tinypng.com) on 2026-11-15. ShortPixel: same four images compressed via the web tool (Lossy/SmartCompress) on 2026-11-15. Both competitors: output bytes measured from downloaded files; SSIM/PSNR via scikit-image (RGB) against the same sources; PNG outputs checked for RGBA-exact (lossless) vs lossy. ShortPixel pricing/features: official pages checked 2026-11-14 (REPORTED). Data: compression-measurements.csv, quality-metrics.csv, network-zero-upload-proof.json, tinypng-skimage.json, shortpixel-skimage.json.