Browser Media Exports: Measured Files, Failed Targets, and Tradeoffs

What these measurements tell you

We prepared two synthetic images and one short generated motion clip, then exported them in Chrome on September 6, 2026. The downloadable files above let you inspect our results and repeat the comparison with your own browser. These are controlled examples, not a claim that every photograph, screenshot, or video compresses by the same amount.

The most useful result was not the smallest file. Re-exporting our 156,833-byte text sample as PNG produced 190,496 bytes: about 21% larger. A JPG at quality 82 was also slightly larger than that source. A tool can encode a file successfully without improving it. Keep the original when the new file offers no useful tradeoff.

Reproduce the image comparison

1. Download the original text sample PNG from the measured-exports section. Start from that source for each run; do not compress an already compressed output.
2. Open Image Compressor, choose WebP, set quality to 82, and keep the original dimensions. Our output was 83,364 bytes at 1600 × 900, about 47% smaller than the source.
3. Inspect small lettering, diagonals, gradients, and the detailed panel at full size. The same quality number is not a guarantee of visually equivalent results across formats.
4. For the smaller-dimension comparison, enable Limit dimensions in Image Compressor and set maximum width 800 and maximum height 450. Export WebP at quality 82. Our direct processing experiment produced 40,052 bytes; browser encoder versions may produce different bytes.
5. Try the 10 KiB target with original dimensions. Our engine tried eight quality settings and stopped at 31,180 bytes. It did not reach 10,240 bytes. Reduce dimensions or revise the target instead of assuming a target input guarantees a matching output.

The downloadable artifacts come from a local browser experiment using the compressor's actual image-processing functions. The experiment bypassed the file-picker and download interface. It verifies the encoding results for these inputs, not every step of the public interface. Encoder versions and intermediate saves can change exact bytes.

Transparency changes the decision

The transparent badge starts at 31,873 bytes. WebP at quality 82 produced 17,808 bytes and retained transparent corners. JPG produced 24,802 bytes with a white background, matching the compressor's JPG setting. Put both over a dark background to see why these files cannot be used interchangeably.

Choose Image Format Converter when the destination requires a specific format. If the graphic must blend into different backgrounds, use an alpha-capable format. JPG has no transparent background to preserve; choose the intended matte color or retain the original.

The badge's 10 KiB target also failed: its smallest result in this run was 12,620 bytes after eight attempts. Transparency and detail still cost space even in a visually simple graphic.

An AVIF option is not evidence of AVIF encoding

Both AVIF attempts failed in this Chrome run. The canvas returned a different format and the processing function rejected that fallback. No AVIF file appears in the downloads because none was produced. Browser image decoding support does not, by itself, establish export support. Try WebP or PNG if your browser reports that an output format is unavailable.

Check the video as a separate tradeoff

Use the generated motion clip and measured presets below to compare resolution, frame rate, file size, and sound. The 440 Hz tone is deliberate test audio; lower your playback volume before listening. The clip is a moving test pattern, not footage of people or a customer screen recording.

Open Video Compressor, start with Web upload, and compare Email attachment or Silent preview from the original file. Silent preview intentionally removes the audio track. The smaller email version reduces the height and frame rate as well as changing encoding quality, so a size difference cannot be attributed to CRF alone.

During this check we found and corrected a resolution mismatch: a control labeled 720p had limited the longest edge to 720 pixels. The corrected control limits height to 720 pixels, preserves aspect ratio with even dimensions, and does not upscale. Check the displayed output dimensions before starting; portrait and landscape files will have different widths.

For a real recording, inspect the duration and dimensions with Video Info Inspector, then remove unused time with Video Trimmer before compressing. Choose a readable frame with Video Thumbnail Extractor. A test-pattern result does not predict text legibility or compression ratio for your recording.

What remains outside this test

We have not measured Safari, Firefox, mobile memory limits, photographs, HDR video, color-profile fidelity, or long recordings in this case study. The browser's CPU and memory budget can make a short successful export a poor predictor for a large file. Keep source files, compare the exported content at its intended display size, and use a controlled production workflow when metadata or color fidelity is essential.

See the Media Publishing Workflow for the order of decisions. The measurements here are evidence for a few specific choices, not a universal quality score or a promised upload size.