# Compression Benchmarks — Real Numbers From Our Tests

> Real compression numbers from our tests: HEIC vs JPEG vs WebP sizes, target-size results for forms and email, measured video compression. Reproducible.

Source: https://toolpantry.app/compression-benchmarks

## About

Every compression figure below was measured on an actual test file using the free tools linked on this page — no estimates, no vendor claims. Rerun any of them yourself in a minute.

## Quick answer

The short version: converting to WebP cut image size 30–60% versus JPEG ; a scanned PDF shrank 85% ; a phone photo reached a 50 KB form limit (94% smaller); a 12-second 1080p video dropped from 17.7 MB to 7.9 MB at 720p and 1.8 MB at 480p.

## Methodology

| Source file | As JPEG | As WebP |
|---|---|---|
| HEIC photo, 29 KB (iPhone) | 46.4 KB | 23.4 KB |
| AI detail image, 1.13 MB | 524 KB | 477 KB |
| Flat illustration, 318 KB | 74 KB | 34 KB |

WebP won every time — most dramatically on flat artwork (−89% versus the original). The HEIC case is interesting: converting *to* JPEG actually grew the file (29 → 46 KB), because HEIC is already an efficient format; the win there is WebP, not JPEG. Try it: [HEIC to JPG](https://toolpantry.app/heic-to-jpg.html) · [HEIC to WebP](https://toolpantry.app/heic-to-webp.html) · [image compressor](https://toolpantry.app/compress-image.html).

| Source file | Target | Result |
|---|---|---|
| Phone photo, 823 KB | Under 50 KB (exam form) | 49.8 KB (−94%) |
| Scanned PDF, 793 KB | Email-friendly | 122 KB (−85%) |
| PDF with removed pages, 812 KB | — | 510 KB (−37%, pages deleted) |

The 50 KB case matters for real forms: Indian exam portals (SSC, UPSC and others) commonly demand photos under 50 KB with specific pixel dimensions. Our [SSC/UPSC size guide](https://toolpantry.app/blog/ssc-upsc-photo-signature-size-2026.html) covers the exact specs; the [compress-to-50KB tool](https://toolpantry.app/compress-image-to-50kb.html) does the work locally.

| Target | Result | Reduction |
|---|---|---|
| Email (keep 1080p) | 15.54 MB | −12% |
| WhatsApp quality | 7.90 MB @ 720p | −55% |
| Discord (under 10 MB) | 8.39 MB @ 720p | −53% |
| Smallest practical | 1.78 MB @ 480p | −90% |

Each re-encode finished in seconds on our test machine. The same clip re-encoded to a looping GIF came out at 320×180, 51 frames, 0.36 MB. Tools: [video compressor](https://toolpantry.app/compress-video.html) · [video to GIF](https://toolpantry.app/video-to-gif.html).

## Why these numbers matter

Most "compression" advice online is either vendor marketing or guesswork. These tests exist because we needed the numbers ourselves — for guides like [the Gmail 25 MB reality check](https://toolpantry.app/blog/gmail-attachment-limit-25mb.html) and [sending video on WhatsApp without ruining it](https://toolpantry.app/blog/send-video-on-whatsapp-hd.html). The full map of platform limits lives in our [file size limits reference](https://toolpantry.app/file-size-limits.html). Every tool runs in your browser — nothing is uploaded — so you can verify any row above with your own files in about a minute.

## FAQ

**How much smaller does WebP make a photo compared to JPEG?**

In our tests, WebP was 30–60% smaller than JPEG at equivalent quality: a HEIC photo converted to 46.4 KB as JPEG but only 23.4 KB as WebP, and a flat illustration went to 74 KB as JPEG versus 34 KB as WebP.

**Can a scanned PDF really be compressed by 85%?**

Yes, when the scan contains image data that can be recompressed. Our test scan went from 793 KB to 122 KB (−85%). Text-only PDFs that are already optimized have much less room — often under 20%.

**How small can a photo be squeezed for a form that asks for under 50 KB?**

A typical phone photo (823 KB in our test) reached 49.8 KB for a 50 KB limit — a 94% reduction that still looked acceptable at form-photo sizes.

**How much does compressing a video reduce the file size?**

In our tests, a 12-second 1080p clip (17.7 MB) dropped to 7.9 MB at 720p (−55%) for WhatsApp-quality output, 8.4 MB for a 10 MB Discord target, and 1.8 MB at 480p (−90%). Results depend on motion and detail in the source.

**Are these numbers reproducible?**

Yes — every number comes from the free tools linked on this page, which run entirely in your browser. Compression results vary with content (flat areas compress far better than detail), so treat these as typical results, not guarantees.