Compress WebP

JPEG divides an image into 8 × 8 blocks and describes each one with a discrete cosine transform. WebP's lossy mode borrows from video coding instead: it predicts each block from the pixels already decoded above and to the left of it, and then stores only the difference between the prediction and reality.

Because most of an image is fairly predictable from its neighbors, that difference is usually small, and small numbers cost few bits. This is why WebP reaches the same apparent quality as JPEG in roughly a quarter fewer bytes, and why it degrades more gracefully at low quality — it blurs rather than breaking into visible 8 × 8 squares.

WebP also carries an alpha channel and a lossless mode that JPEG has never had, which is what lets one format replace both JPEG and PNG on a web page.

Lossless compression rewrites the file so every pixel survives exactly — decode it and you get the original back, byte for byte. PNG, GIF, lossless WebP and ZIP all work this way.

Lossy compression discards detail on purpose. JPEG converts the image to a frequency representation and throws away the high-frequency detail your eye is least likely to miss. That is irreversible, and it is how lossy reaches sizes lossless cannot: a photograph saved as PNG is routinely five to ten times larger than the same photograph as JPEG.

Neither is better. Photographs suit lossy encoding; screenshots, logos and line art suit lossless, because lossy encoders leave haloes around hard edges.

The quality number is not a percentage. It is an index into quantization tables that control how aggressively detail is discarded, and its relationship to file size is steep and non-linear.

• 95–100 — near-original, often only 20–35% smaller. Rarely worth it.
• 80–90 — photography you care about. 50–70% smaller, no artifact at normal viewing distance.
• 70–80 — ordinary web images. Usually 70–85% smaller.
• 50–70 — visible softening in smooth areas; thumbnails survive it.
• Below 50 — blocking, ringing and color banding. Placeholders only.

Returns diminish fast at the top: 100 to 90 usually halves the file for no perceptible change, 90 to 80 saves far less. If you remember one number, use 80.

JPEG has no alpha channel and cannot store transparency. Converting a transparent PNG forces those pixels to become something, and most tools fill them with black — which is why a logo exported to JPEG arrives in an unwanted dark box. Keep it as PNG, use WebP (alpha support, typically 25–35% smaller than PNG), or flatten onto a chosen background deliberately.

Color is the other gap. Screens mix additively in RGB; printing presses mix subtractively in CMYK, and converting between them always approximates. A CMYK JPEG is converted to RGB before a browser can display it, and saturated cyans and deep blues can shift noticeably. Compress and resize in RGB for screens, and keep a separate CMYK master for print.

Every time a JPEG is decoded, edited and saved, the encoder re-runs quantization on data already quantized. Errors from the previous pass are treated as real detail and collect new errors on top. This is generation loss, and it only accumulates.

Re-saving at quality 90 restores nothing; it spends more bytes preserving the earlier pass's artifacts. Compress from the highest-quality original you have, and settle on one output size in a single pass rather than nudging quality down and re-saving.