benchmark · AVIF quality benchmark

AVIF Quality Benchmark

AVIF can become very small, but aggressive quality settings need careful review.

Benchmark notes

  • Quality 50 reached 133 KB.
  • Quality 40 dropped sharply to 40 KB.
  • Quality 30 was tiny but mainly useful as a stress test.

AVIF quality results

QualityOutput sizeChange from q70Useful first testCaution
70380 KBBaselineHigh quality AVIFSupport must be checked
60287 KB-24%Balanced AVIFCheck browser and CMS
50133 KB-65%Small web imageReview texture
4040 KB-89%Aggressive targetVisible loss likely
306 KB-98%Stress test onlyToo low for many photos

Test methodology

ImgShrink generated four repeatable samples with sharp: a photo-style gradient, a UI screenshot, a transparent graphic and a flat color illustration.

Each output was encoded with the stated format and quality setting, then measured by byte size before being rounded for the table.

Test images and parameters

Photo sample: 1600x1067 pixels. Screenshot sample: 1365x768 pixels. Transparent sample: 512x512 pixels. Flat graphic sample: 1200x800 pixels.

WebP used quality 78 for the cross-format table, AVIF used quality 50, and JPEG baselines used quality 90 unless a quality row states otherwise.

What we learned

Quality 50 reached 133 KB.

Quality 40 dropped sharply to 40 KB.

Quality 30 was tiny but mainly useful as a stress test.

Limitations

These values are benchmark results for the sample set, not a promise for every uploaded file.

Real images vary by noise, detail, transparency, dimensions, browser encoder and the target platform's accepted formats.

Benchmark questions

Are these exact results for my image?

No. They are measured sample results that help choose a first setting to test.

Do these tests upload user images?

No. They use generated samples and browser-style encoding assumptions.