Skip to main content
GARGLOO

Guide · 5 min read

How to Reduce Image File Size

Make photos and screenshots smaller for the web, email and forms: quality settings, resizing and format choice explained.

Last updated: 2026-10-07

Tools for this guide

The short answer

To make an image file smaller, resize it to the dimensions you actually need, then lower the quality to around 80%, and save photos as JPG and web graphics as WebP. Resizing usually does most of the work; quality and format choice do the rest.

On Gargloo this maps to three free tools used in order: Image Resizer for dimensions, Image Compressor for quality or exact size targets, and the JPG, PNG and WebP converters for format changes. Everything runs in your browser, so your originals are never uploaded anywhere.

Resize dimensions first

File size scales with pixel count, not just quality. A phone photo at 4032×3024 shown in an 800px-wide article carries about 25 times more pixels than displayed.

Use the Image Resizer to match the actual display size (plus a little margin for high-density screens, typically 2x at most).

Then tune quality

The Image Compressor's quality slider is the main lever for JPG and WebP. Detailed photos shrink a lot at balanced quality; screenshots with flat colors shrink less because there's less to discard.

Always compare the before/after preview at full zoom before downloading — banding in skies and blurry small text are the first warning signs you've gone too far.

Pick the right format

Photos: JPG or WebP. Graphics with transparency: PNG or WebP. WebP files are usually clearly smaller than equivalent JPG — see our JPG vs PNG vs WebP guide for the full decision tree.

PNG output from a compressor helps less than you'd expect, because PNG is lossless; converting photographic PNGs to WebP is usually the bigger win.

Understand Quality mode and Target size mode

The Image Compressor offers two working styles behind a single mode selector, and choosing the right one shapes the whole session. Quality mode hands you direct control: pick an output format, move the quality slider between 30 and 100 percent, optionally cap the width, and every file is encoded exactly as specified. The hint beneath the controls narrates the tradeoff as you drag — high values bring modest savings, middle values usually look indistinguishable from the original, and aggressive values warn of visible artifacts on detailed areas. Use this mode when visual consistency matters more than an exact byte count.

Target size mode reverses the relationship: declare the file size you need and the tool adjusts quality and dimensions on its own until each file lands near the goal. It reads real pixel dimensions first so the adjustment ladder can downscale when quality steps alone cannot reach the target, which also covers PNG output that has no quality dial to turn. When a file cannot honestly reach the goal, the result note reports the closest achievable size and keeps the smallest reasonable output instead of degrading the image into mush. Use this mode when a form or inbox states a cap and the bytes matter more than matching a particular quality number.

Set output format deliberately after sizing

Dimensions fix the pixel count while format fixes the cost per pixel, so decide format after the size is right rather than before. The compressor accepts JPG, PNG, and WebP inputs and offers JPG, WebP, and PNG outputs: JPG for photographs headed to forms, marketplaces, and sharing pipelines that accept it everywhere; WebP for web publishing where leaner files pay off on every page view; PNG when exact pixels matter, such as logos, line diagrams, and screenshots with small text. The default quality sits at 82 percent, a balanced starting point for photographic content that you can raise for portfolios or lower for drafts.

Read the interface note when PNG is selected — it states plainly that PNG stays lossless and suggests JPG or WebP for smaller files. Treat that note as routing rather than decoration, because photographic content saved as PNG can weigh several times what the same picture costs as JPG. Switching a supplier-provided photographic PNG to WebP after resizing typically resolves size problems that quality adjustments alone cannot touch. For mixed collections, sort by content before configuring: photographs in one run, graphics in another, since a single format rarely flatters both kinds at once.

Use limit width to tame oversized photos

Phone cameras record far more pixels than most destinations can display, and the optional limit-width field in Quality mode converts that surplus into savings without touching quality. Type a value such as 1920 and any image wider than that threshold is scaled down proportionally while smaller images pass through at native dimensions. Because the scaling preserves aspect ratio automatically, a landscape and a portrait sharing one width cap each keep their own correct heights. The operation runs inside the same encoding pass as compression, so no second step is required.

Choose the cap from the destination rather than habit: article illustrations rarely need more than the column width plus a margin for high-density screens, profile uploads should match the stated pixel requirement, and email attachments benefit from widths that keep the long edge near screen size. Values must fall between 16 and 20000 pixels to be accepted. When a form demands exact dimensions rather than a maximum edge, switch to the Image Resizer instead, which controls width and height explicitly with aspect locking and named presets. Limit width is a ceiling for casual sharing; exact dimensions are a specification for strict gates.

Handle PNG files that refuse to shrink

A photographic PNG that barely moves under compression surprises many uploaders, yet the behavior follows directly from the format design. PNG encodes pixels without loss, so the quality slider cannot apply and the tool disables it with an explanatory note instead of pretending. Screenshots with flat colors and sharp edges compress reasonably under PNG because large uniform areas encode cheaply, but camera photographs with grain, gradients, and foliage carry entropy that lossless encoding cannot discard. Running such files through repeated PNG passes wastes sessions while the byte count stands still.

The remedy is conversion rather than stronger settings. Re-encode photographic PNGs as JPG for universal acceptance or as WebP for leaner web delivery, starting near the default quality and comparing the before-and-after preview at full zoom. Reserve PNG output for content that genuinely needs it: transparency, crisp lettering, interface captures, and archival intermediates headed for further editing. When a strict portal mandates PNG regardless of content, use Target size mode, which shrinks dimensions instead of quality for lossless output — and verify that the reduced pixels still satisfy the form before submitting.

Compress batches without repeating setup

Real collections arrive in groups — product catalogs, event albums, documentation screenshots — and the compressor queues up to 10 files per run at up to 60 MB each so a coherent set shares one configuration. Add the files together, confirm the format and mode once, and start the run; progress is reported per file with a cancel control that halts the batch if a wrong setting surfaces mid-run. Shared settings keep results visually consistent, which matters when images belong to the same gallery or listing and uneven quality would show.

Organize before queuing rather than during. Pull outliers aside — a transparent logo among photographs, a single graphic with text among landscapes — and run them separately with settings suited to their content instead of compromising every file to a middle value. Keep originals untouched, since compression discards data and a downscaled copy cannot regenerate the detail it shed. When one file in a batch disappoints, re-run that file alone from its original with adjusted quality or format rather than reprocessing the finished set.

Preview at full zoom before downloading

A compressed image deserves a short inspection before it ships to a storefront, hiring portal, or printed page. Open the downloaded copy at 100 percent and study the regions where damage appears first: skies and smooth gradients for banding, foliage and fabric for smearing, faces for waxy smoothing, and any baked-in lettering for softened edges. Compare identical regions against the original at matching zoom rather than trusting thumbnails, which hide exactly the defects that full-size viewing reveals. The before-and-after preview inside the tool serves the same purpose before anything is saved.

Make corrections systematically by changing one variable per re-run so each comparison teaches what moved the result. Banding across gentle tones means quality fell too far for that content — step gentler or switch formats. Soft lettering means dimensions dropped below what the content needs — no quality value rescues text rendered at half its required pixels. Mottled textures point to repeated encoding from an already-processed copy, and the cure is returning to the master and running once with corrected settings. Log the winning combination for recurring content types so future batches start from a proven recipe.

Prepare images for email, forms, and web pages

Each destination judges images by different rules, so configure for the gate in front of you. Email attachments care about total message weight: resize photographs toward screen dimensions, encode as JPG near balanced quality, and confirm the combined attachments sit below the stated cap with margin. Upload forms check format, dimensions, and file size together: match the required pixels first with the resizer, then address bytes with target mode, verifying both numbers in file properties before submitting. Web pages reward every kilobyte saved across all visitors, so WebP output with thoughtful widths compounds into faster loads and lower bandwidth.

Test one file through the real destination before committing a batch. Upload a single compressed copy through the actual form, attach it to a draft message, or preview it on a staging page, and confirm it renders, submits, and displays correctly. Format rejections mean the pipeline wants a different container; dimension rejections mean re-checking width against height order; size rejections mean compressing from the verified copy rather than lowering quality blindly. Only after the pilot file clears every gate should the remaining files follow the same path from their originals.

Recover from soft or banded results

Failed outputs show recognizable symptoms and each symptom prescribes its own correction. Visible stair-steps across skies and studio backdrops signal that quality dropped too far for gentle tonal transitions; re-run from the original one step gentler or move photographic content out of lossless workflows that cannot help. Softness in captions and label edges signals over-compression or a resize below content needs; confirm dimensions first, then raise quality and compare side by side. Smearing in detailed textures after several passes signals cumulative encoding damage — return to the source file and run once instead of polishing a processed copy.

Separate casual sharing from strict submissions when judging recovery. A slightly aggressive file may still serve a chat thread or internal draft where speed matters more than pixel perfection, while the same file would fail a product listing or print order that invites close inspection. Keep both the master and the delivery copy with clear names so each future use starts from the right source. Over time, note which content types tolerate which settings: bright daylight scenes forgive more than dim grainy interiors, and graphics with text forgive less than anything. That record turns guesswork into routine.

Coordinate compression with cropping and rotation

Compression sits mid-pipeline among companion operations, and sequencing them correctly preserves quality that wrong ordering squanders. Crop first with the Image Cropper to remove desk edges, excess sky, and unwanted borders before anything else, because deleted pixels never consume encoding budget and composition corrections stay sharpest at full resolution. Straighten horizons and rotate sideways captures at this stage as well, while pixel density is highest. Only then compress or resize the cleaned frame, so every subsequent operation works on intended content rather than discarded surroundings.

Treat each correction as a generation and minimize their count ruthlessly. Every encode discards data, so a file cropped, resized, compressed, converted, and recompressed accumulates five rounds of approximation where two careful passes would have sufficed. Plan the chain on paper for important images: crop dimensions, final size, output format, then execute once per stage from the strongest available predecessor. When settings need revision, return to the master and rerun the chain rather than patching the finished file. Fewer generations with correct parameters always surpass many generations of fixes.

Serve responsive variants instead of one middle file

Web pages display identical images across phones, tablets, laptops, and wide monitors, and a single compromise file serves all of them poorly — too heavy for mobile data, too soft for large screens. Generate responsive variants from the archived master instead: a modest width for feeds and cards, a medium width for article bodies, and a full width for heroes and lightboxes, each compressed appropriately for its role. Thumbnails at 256 pixels need far less fidelity than showcase images, so tune quality per variant rather than applying one value everywhere. Modern markup selects among variants automatically per viewport.

Name variants systematically so templates reference them unambiguously across redesigns: portrait-card-800.jpg beside portrait-hero-1920.jpg prevents the midnight mistake of serving a thumbnail to a fullscreen slot. Verify each variant at its display size rather than only at full zoom, since downscaled rendering hides flaws that magnification exaggerates — judge fitness for purpose, not abstract perfection. Measure the page weight saved across the variant set to confirm the effort paid off, and regenerate the whole family from masters whenever art direction changes instead of stretching old files into new slots.

Protect originals with an archiving routine

Every compression session should end with files in three distinct places serving three distinct roles. Masters live in a dated archive hierarchy at full resolution, retaining metadata and quality for future regeneration across unknown requirements. Working copies occupy a staging folder during active preparation, where cleaning, resizing, and format experiments happen without endangering sources. Deliveries reside in per-purpose folders named for their destinations, carrying the exact pixels, formats, and sizes each gate demands. No file should serve two roles simultaneously, since role confusion is how originals get overwritten by derivatives.

Enforce the separation with naming and habit rather than memory. Mark masters distinctly so they never attach to messages by accident, date deliveries so resends reference the correct variant, and record the settings behind recurring deliveries for reproducibility. Back up the masters folder on a schedule independent of any single device, because device loss without backup destroys the regeneration capability that makes every other practice work. Compression is reversible in workflow terms only while masters survive; without them, each delivery is a dead end rather than a renewable starting point.

Device notes: iPhone, Android, Windows, Mac

On iPhone, photos arrive as HEIC files that many forms reject: convert them to JPG with the HEIC to JPG tool first, then resize and compress. Move images through the Files app — save from Photos, then drop them into the tool from Files.

On Android, pick images straight from the gallery or the Files app. Phone screenshots are far larger than any screen they will be shown on, so resize them before compressing and the savings are immediate.

On Windows, drag files straight from File Explorer into the tool window. Screenshots from the Snipping Tool compress especially well once resized to their display size.

On Mac, screenshots land on the Desktop at double density by default, so resizing does most of the work before quality settings matter. Photos exported from Preview as PNG should usually become JPG or WebP first when file size is the goal.

Frequently asked questions

What quality setting should I use?

Around 80% is visually indistinguishable from the original for most photos. Drop to 60–70% for thumbnails and previews where nobody will zoom.

Should I resize or just compress?

Do both when the dimensions are larger than needed: a 4000px photo displayed at 800px wastes roughly 25x the pixels. Resize first, then compress.

Why did my PNG barely shrink?

PNG is lossless, so the encoder preserves every pixel exactly and quality adjustments do not apply — the tool disables the quality slider for PNG and says so directly. For photographic PNGs, switch output to JPG or WebP, which discard fine detail in exchange for much smaller files.

Can I hit an exact file size automatically?

Yes. Flip the compressor from Quality to Target size mode and declare the size you need. The tool then adjusts quality and dimensions on its own until each file lands near the goal, and reports honestly when a file cannot reach it instead of destroying quality.

Try it now — free, no signup

Batch compression with exact size targets. Files stay on your device.

Open Image Compressor
How to Reduce Image File Size | GARGLOO