How to compress JPG, PNG, WebP, GIF, SVG, and AVIF images

Start by identifying where the image will be used and the maximum dimensions the destination actually displays. Keep an untouched original, crop empty or irrelevant areas, and remove duplicates from the batch. When a 4000-pixel photograph will appear as a 1200-pixel website image, resizing before or during compression removes data the viewer could never see. Upload the working copy and download the optimized result under a name that distinguishes it from the master.

Judge the result at the intended display size first, then zoom in to inspect vulnerable detail. Read screenshot text, examine faces and product edges, and test QR codes or barcodes with a real device. Check transparency when using PNG or WebP, and confirm that animated files still animate if that behavior matters. Record both pixel dimensions and file size; an image can have a small download size but still waste layout space if its dimensions are unnecessarily large.

Lossless compression reorganizes data so the original pixels can be reconstructed exactly. It works well for flat graphics, line art, interface captures, and images where every hard edge matters, but it cannot always produce tiny files. Lossy compression discards visual information that an encoder predicts people are less likely to notice. It is especially effective on photographs, gradients, and textured scenes, although strong settings can create blockiness, halos, banding, or smeared detail.

The word lossless does not automatically mean better for every purpose. A lossless photograph may be many times larger than a visually similar lossy version, slowing a page or failing an upload limit. Likewise, a heavily compressed JPG screenshot can make text unpleasant to read. The useful choice is the method that preserves the information the image must communicate while staying within the performance, storage, or submission budget.

JPG is a dependable choice for ordinary photographs and broad compatibility. PNG preserves transparency and sharp flat-color detail, making it useful for screenshots and graphics, but photo-like PNG files can be large. WebP can support lossy or lossless images and transparency with good web efficiency. AVIF can achieve excellent photo compression, though older software may not accept it. GIF is widely recognized for simple animation but has limited color, while SVG stores vector shapes and text rather than a fixed pixel grid.

Do not change format merely because one format is newer. The receiving portal, content-management system, email client, or design application must support the output. Logos and diagrams often belong in SVG when the workflow permits it; photos usually do not. Screenshots with tiny interface text favor PNG or a carefully tested modern format. When compatibility is uncertain, keep a well-optimized JPG or PNG fallback and verify the actual upload rather than relying on the filename extension.

Image compression depends heavily on the file type. JPG and WebP usually shrink well because they can reduce visual detail. PNG files with flat colors may already be efficient, while PNG photos can be much larger than needed.

SVG files are text-based vector images, so compression often means cleaning metadata rather than changing pixels. GIF and AVIF files have their own limits, especially when animation or advanced color data is involved.

Crop empty background, remove duplicate screenshots, and resize oversized photos when exact dimensions are not required. A 4000 pixel photo used in a small form field wastes upload space.

Keep the original image if it is needed for printing, brand work, or future editing. Use the compressed output for sharing, uploading, or placing inside documents.

After download, zoom in on text, faces, labels, QR codes, barcodes, and fine lines. These details are usually the first things to suffer when compression is too aggressive.

For website use, compare file size and visual quality together. The best image is not always the smallest one; it is the smallest file that still looks appropriate in its real context.

When compressing many images, group similar files together so you can review results faster. Product photos, scans, and screenshots often need different quality expectations.

Uploaded images are processed for the compression request you start and are intended to be removed from temporary storage after processing. Avoid uploading images that your policy does not allow online.

The steps below match how Compress Image on PDF Convert Now currently behaves. Accepted input: Image files. Up to 20 files, 50 MB total.

Example: Compress a photo set for web upload and separately validate one screenshot where small text and icon edges must remain sharp.

When the tool returns a generic download name, it uses compressed-images.zip; rename it to something descriptive before sharing. Keep the original source file until the recipient or portal accepts the result.

Pixel dimensions often influence size more than small quality adjustments. Halving both width and height reduces the pixel count to one quarter, which can create a substantial saving when the destination never uses the extra resolution. Choose dimensions based on the largest real display size and account for high-density screens when appropriate. Avoid enlarging a small image before compression; interpolation creates more pixels but does not restore missing detail.

For websites, export dedicated thumbnail and full-size versions instead of serving one enormous image everywhere. For online forms, follow any minimum resolution requirements for identity documents or evidence photographs. For print, work backward from physical dimensions and required print quality. A delivery copy for a web page and a print master are different assets, so label them clearly and preserve the master outside automated optimization workflows.

A useful web image loads quickly without looking visibly damaged in its component. Test on a phone-sized viewport and a larger screen, because artifacts can be more obvious when an image is stretched. Avoid placing essential text inside a compressed photograph when real HTML text can do the job more accessibly. Use meaningful alternative text in the page rather than stuffing filenames or alt attributes with keywords.

For document portals, confirm file type, dimensions, size, orientation, and legibility after optimization. Codes, account numbers, fine diagrams, and evidence details deserve closer inspection than decorative backgrounds. If the platform recompresses uploads again, begin with a clean, moderately optimized source instead of an already damaged file. The smallest successful file is the one that passes the destination's rules and still communicates its purpose.

For photographs, begin with cropping and dimensions, then adjust lossy quality while watching faces, hair, fabric, foliage, and smooth skies. Those areas reveal smearing and banding quickly. Product photos also need clean edges, accurate color, and readable labels. Export from the original rather than a social-media download, which may already be compressed. Keep a larger master for marketplaces or print and create purpose-specific web sizes so a thumbnail, product page, and zoom view do not all load the same oversized asset.

For screenshots and interface documentation, preserve small text, one-pixel lines, and flat-color boundaries. PNG or lossless WebP can be more suitable than JPG, but cropping unused browser chrome and scaling to the real documentation width often matters most. Avoid shrinking until text becomes unreadable, and do not use an image as the only source of instructions. For diagrams and logos, obtain SVG or another vector source when possible; compressing a raster screenshot of vector artwork locks it to a limited resolution and makes future reuse harder.

For scans, IDs, receipts, and QR or barcode images, legibility and machine recognition outrank decorative quality. Keep the code large enough to scan, preserve contrast around it, and test the actual downloaded output. Do not crop security features or required borders from identity documents. Batch workflows should separate these functional images from ordinary photos so a single global setting does not damage critical detail. Keep a small sample set representing the hardest content and compare it whenever a workflow, encoder, or upload platform changes.

Build a small acceptance checklist for repeatable image work. Define the maximum display dimensions, preferred and fallback formats, target size range, color or transparency needs, and details that must remain legible. Test on a representative slow connection or constrained upload when performance is the goal. Store the master, export settings, and optimized derivative separately, and avoid stripping metadata that a rights or production workflow actually needs. Conversely, review location, device, or author metadata before public sharing when privacy matters. Compression quality is not a single slider value: it is the combined result of source quality, crop, dimensions, format, encoder, destination rendering, and the viewer's real task.