How to choose an image quality setting
Use a repeatable visual review to balance file size against detail instead of treating the quality control as a percentage score.
An encoder quality value is an instruction to a particular codec, not a promise that 82 means 82% of the original quality. The same number can behave differently across JPEG, WebP, and AVIF and across images with different texture, noise, text, or gradients.
Quality is codec-specific
Lossy codecs divide and simplify visual information in different ways. A JPEG quality value cannot be compared directly with the same WebP or AVIF value. Even two encoders for the same format may make different decisions at the same setting.
Ziply begins at quality 82 as a practical review point for lossy output. It is not a universal optimum. A clean studio photograph, a noisy phone image, and a screenshot with small text can require different settings even when their dimensions match.
Review the regions that fail first
Compression artifacts are easier to see around sharp contrast changes and in areas that should vary smoothly. Zoom in on eyes and hair in portraits, leaves and grass in landscapes, lettering in screenshots, and gradients in sky or product backgrounds. Repeated textures can become smeared or develop blocks before the overall image looks obviously degraded.
Then zoom back out. A tiny artifact visible only under extreme magnification may not matter for a thumbnail, while a subtle halo can be distracting in a full-width banner. The delivery size and viewing distance belong in the decision.
Use a short candidate ladder
Instead of moving the control one point at a time, create a few meaningfully different candidates. Start at the default, try a lower setting, and keep lowering until a visible defect appears. Return to the smallest candidate before that threshold.
For a large batch, test several difficult representatives first. A setting that works on the easiest photo may fail on the image with fine text or dense texture. Splitting mixed content into separate batches can produce better results than forcing one value across everything.
- Compare the original and output using the same crop and zoom.
- Record both visual result and byte size.
- Test the destination’s resizing or recompression behavior.
- Retain the source so a future codec can be generated without cumulative loss.
Know when resizing matters more
Pixel dimensions often dominate the byte budget. A 4000-pixel image displayed at 800 pixels can waste more transfer than a modest quality adjustment can recover. Premium maximum-dimension controls can create the appropriately sized local delivery copy before encoding.
Do not downsize the only master. Export for the real layout, keep enough resolution for high-density screens when needed, and preserve the original separately. This produces a predictable delivery asset without turning a reversible optimization step into permanent data loss.