How to Compress Images Without Losing Quality

Lossless vs. Visually Lossless

When people search for compression "without losing quality," they usually mean one of two things, and the distinction decides which format you reach for.

The trap: lossless compression barely helps photographs.Photos are dense with sensor noise and smooth gradients — exactly the content that lossless encoders can't find patterns in. A PNG of a photo is often larger than the JPG it came from. For photographic content, visually lossless is the only route to a meaningfully smaller file. Lossless is for the other kind of image: flat color, hard edges, text, transparency.

The Best Image Compression Format

Format choice moves file size far more than any quality slider. These are the four that matter, compared against JPG as the baseline everyone knows:

Pick WebP if you pick one thing. It compresses both ways, handles transparency and animation, and decodes in roughly 97% of browsers — you can drop it into a plain <img> tag without a fallback.

Add AVIF if you serve a lot of photos.It's another 20–30% smaller than WebP, but it encodes 5–20× slower and support sits near 94%, so in production it's served as the first source in a <picture> element with WebP behind it. Our WebP vs. AVIF comparison covers the trade-offs and the exact fallback markup.

Quality Settings That Stay Invisible

Once you've chosen a lossy format, the quality slider decides everything. These bands hold for JPG and lossy WebP, whose scales are broadly comparable:

• 95–100% — near-transparent compression, but file size climbs steeply for detail nobody perceives. Rarely worth it.
• 85–90% — the visually lossless band. Roughly half the size of a 100% export with no perceptible difference. This is the setting you want.
• 75–80% — aggressive but usable for thumbnails and feed images. Artifacts appear on close inspection around hard edges.
• Below 70% — visible softness and blocking on most photos. Only when bandwidth genuinely trumps fidelity.

AVIF's numbers don't map onto this scale. Most AVIF encoders reach the same perceptual quality around 50–65, so copying "90" across from your JPG workflow produces a needlessly large file. Judge AVIF by eye, not by the number.

Compression Wins That Cost No Quality At All

Before touching a quality slider, take the free savings. Each of these shrinks the file without discarding a single visible pixel:

• Resize to display size. The single largest win. A 6000px photo shown in a 1600px column carries ~14× more pixels than the screen can render. Downscale first — the result is both smaller and sharper than the full-size image crushed to the same byte count.
• Strip metadata. Camera EXIF, GPS coordinates, embedded thumbnails, and color profiles can add tens of kilobytes to a small image. Most converters drop them by default. (Do keep the color profile if the image is wide-gamut.)
• Run a lossless optimizer. Tools like oxipng for PNG and jpegtran for JPG re-pack the existing data more efficiently — same pixels, smaller file, no re-encoding.
• Reconsider the format. Converting a screenshot from PNG to lossless WebP is a free ~25% reduction with pixel-identical output. No slider involved.
• Use SVG where it applies. Logos, icons, and diagrams that were drawn as vectors should ship as vectors — infinitely scalable and usually a fraction of the size of any raster export.

The Mistakes That Actually Destroy Quality

• Compressing an already-compressed file. Every lossy re-encode re-quantizes the image. JPG → WebP → JPG compounds the artifacts even at high quality settings, and the damage is permanent. This is generation loss. Always start from the original.
• Discarding the master.Compress a copy. When a better format arrives — as WebP did, as AVIF did — you re-export from the lossless original, not from last year's JPG.
• Saving a photo as PNG to "keep quality." It keeps quality, and it also produces a file several times larger than a visually lossless WebP. Lossless is the wrong tool for photographic content.
• Compressing a transparent PNG into JPG. JPG has no alpha channel, so transparency is flattened onto a solid background — usually white or black. Use WebP or AVIF, both of which compress and keep the alpha channel.
• Upscaling before compressing. Enlarging an image invents pixels the encoder then has to spend bytes on. It never adds detail.
• Judging quality on a thumbnail.Always inspect at 100% zoom, on the content that matters — skin tones, gradients, text edges. That's where lossy compression shows itself first.

Which Tool to Use

A Workflow That Works Every Time

1. Work on a copy. Keep the original master untouched.
2. Resize to the largest dimensions the image will ever be shown at.
3. Choose the format by content: photo → lossy WebP or AVIF; graphic, screenshot, or transparency → lossless WebP.
4. Set quality to 85–90% for WebP or JPG, 50–65% for AVIF. Skip this entirely for lossless.
5. Check one output at 100% zoom against the original before committing to the rest.

Doing this to a whole folder at once is the same five steps with one extra concern — a bad setting silently degrades every file. Our batch conversion guide covers how to test a batch before you run it.