Compress Images to a Target Size

Open the Bulk Image Compressor to apply a per-image budget to still JPG, PNG or WebP inputs. This guide follows a fictional client asset delivery, with invented candidate sizes used to explain the arithmetic. They illustrate the method, not expected output sizes or a benchmark for any particular browser.

Turn a client request into a precise budget

A client request such as keep the images small leaves several questions unanswered. Does the allowance apply to each image, the whole delivery folder, or an email attachment? Does it include supporting documents? Must dimensions stay unchanged? Which formats will the recipient actually use? Resolve those questions before processing. Otherwise, you can produce a tidy batch that follows your interpretation while failing the delivery system or forcing the recipient to ask for everything again.

In target mode, this tool uses decimal KB: one KB is 1000 bytes. A target of 240 KB therefore means 240000 bytes per image. It is not 240 times 1024 bytes, and it is not an amount shared across the batch. File managers can display rounded sizes or use different unit conventions. Use the actual byte figure when a boundary matters instead of comparing two rounded labels that happen to look equal.

The target field accepts values from 10 to 10000 KB and starts at 200 KB. Those are compressor settings, not official allowances for a social network, client portal or messaging service. Enter the budget that belongs to your brief within the available range. If a receiving system has a strict boundary, confirm its unit and acceptance rule directly. Do not invent a universal safety margin and describe it as a requirement for everyone.

For this worked delivery, a fictional visitor-center client requests three JPG review assets at no more than 240000 bytes each. The images must retain their current oriented dimensions for the first review. One is a reception photograph, one is a directions screenshot and one is a simple opening-hours banner. The client has specified a per-file allowance, so success means three independently acceptable files, not merely a small combined total.

Separate the byte rule from the visual acceptance rule

Define what a reviewer must still be able to see. In the reception photograph, that may be the entrance sign and the counter layout. In the directions screenshot, it may be street names and a route legend. In the banner, it may be the exact opening times. These details provide a visual acceptance checklist. The target status supplies a separate byte check, and neither should be used as a substitute for the other.

A quality setting does not express a known percentage of information retained. It requests an encoding level for JPG or WebP, with the actual result depending on the image and browser encoder. A photograph and a graphic at the same quality can have different file sizes and different defects. Record the selected setting to explain the run, but approve the visible result rather than treating its number as a certificate of image quality.

Fixed-quality mode and target mode serve different starting points. Fixed quality lets you choose a whole-number setting from 20 to 95, with 85 as the default, then inspect the resulting bytes. Target mode starts from the byte allowance and selects among its quality samples. Neither removes the need for inspection. If an otherwise acceptable fixed-quality export already meets the client brief, a second search is not automatically a better creative decision.

Keep the original source out of repeated compression loops. If a low-quality review copy loses a street label, increasing the setting on that copy cannot recreate the lost detail. Return to the stored source for each new trial. The compressor does not overwrite your originals, but it also does not save a resumable project or autosave the batch. Your own source folder remains the dependable starting point for later changes.

Read the quality samples as a limited search

For JPG and WebP target mode, the sampled quality settings are 95, 90, 85, 80, 75, 70, 65, 60, 55, 50, 45, 40, 35, 30, 25 and 20. The selection rule is the highest sampled quality whose encoded bytes fit the budget. It is not a test of every possible setting. Keep the word sampled in mind when explaining why a particular candidate was selected for a client delivery.

Suppose the reception photo starts at 1440000 bytes. For this fictional example, quality 95 produces 364000 bytes, 90 produces 302500, 85 produces 251800, and 80 produces 239400. With a 240000-byte allowance, 80 is the highest sampled quality that fits. It leaves 600 bytes unused. The result is within budget; it does not need to equal the target exactly to count as a fit.

The search does not tell us what quality 81, 82, 83 or 84 would produce because those settings are not sampled. It also does not establish that quality 80 has the best perceived detail among every conceivable encoding. You can investigate a different fixed setting separately if that serves the brief, but describe it as another trial. Do not rename the sampled result an optimal compression or a lossless match to the original.

A useful review note for the photo would say JPG, original oriented dimensions, target 240000 bytes, selected sampled quality 80, output 239400 bytes, visual review pending. That record separates settings, arithmetic and approval. A colleague can understand what happened without assuming the visual review was automatic. Return to the image compressor to inspect a real source with the same decision structure, using its actual output rather than these illustrative figures.

Interpret a miss without disguising the result

Now consider the directions screenshot, whose source is 620000 bytes. In the invented sample set, every candidate exceeds the budget and the smallest is 258200 bytes. The compressor keeps that smallest sampled candidate and marks target missed. It is 18200 bytes over the 240000-byte allowance. A substantial reduction from the source does not make it a fit. The delivery rule compares output with budget, not output with its own starting size.

Do not assume the lowest quality setting must always be the smallest candidate in every encoding situation. The defined fallback is the smallest of the samples actually produced. Read the reported result rather than calculating an imaginary output from the quality number. The fallback also does not mean the image is visually acceptable. The screenshot could both exceed the byte budget and have labels that no longer survive close inspection.

A miss is useful information about this source, format, dimension choice and budget. Ask which condition can change. The client might allow a larger reference image, accept another format, or approve a simplified source graphic. Those are decisions to obtain, not changes the compressor can infer. If no condition can change and no acceptable candidate fits, record the asset as unresolved rather than attaching it as though the request had been met.

An output can also fit while growing larger than its source. The opening-hours banner begins at 84000 bytes and, in this example, its quality-95 candidate is 91000 bytes. That candidate meets the 240000-byte budget but adds 7000 bytes. The tool does not secretly replace it with the 84000-byte original. Decide separately whether a new derivative is needed at all, and keep any unchanged original outside the compressor output package.

Use arithmetic that matches the delivery agreement

For the photo, the byte reduction is 1440000 minus 239400, which equals 1200600 bytes. Divide 1200600 by 1440000 and multiply by 100 to obtain approximately 83.38 percent reduction. That figure describes this invented pair of files only. It does not predict another photograph, certify equivalent detail or tell the client whether the entrance sign remains readable. Keep size reduction and suitability as separate lines in your review.

For the screenshot, 258200 minus 240000 equals an 18200-byte overage, or about 7.58 percent of the requested budget. This denominator is the budget, not the source size. For the banner, 91000 minus 84000 equals 7000 bytes of growth, approximately 8.33 percent of its source. Stating what a percentage is relative to prevents a positive-looking number from concealing whether you mean saving, growth or distance above the target.

The three illustrative outputs total 588600 bytes: 239400 plus 258200 plus 91000. Three allowances of 240000 bytes total 720000 bytes, so the output sum is below that sum of allowances. Nevertheless, the screenshot still fails its individual requirement. The banner’s unused space cannot compensate for the screenshot. This is why an aggregate saving is a useful overview but not a complete acceptance check for a per-image delivery contract.

The ZIP adds packaging and includes the CSV summary and README, so its byte size is a different quantity again. The PDF report is another file, not part of an image’s target calculation. If the client also has a total transfer allowance, check the final package separately. Read the exported per-file figures and the actual package properties instead of assuming the sum of image bytes equals the complete attachment size.

Change dimensions only as an explicit second decision

The target search never reduces dimensions to make a candidate fit. With the optional maximum edge set to 0, original oriented dimensions are retained. You can instead choose a longest-edge limit from 640 to 4096 pixels. That choice is applied independently of the sampled quality search. It changes the prepared image you are encoding, so treat it as a new trial with a different visual requirement, not an invisible continuation of the same test.

Suppose the screenshot is 2400 by 1800 pixels and the client later approves a maximum edge of 1600. Its resulting dimensions become 1600 by 1200, keeping the aspect ratio without cropping. The pixel count changes from 4320000 to 1920000. That does not prove its file bytes will fall by the same fraction. Image content and encoding still matter, and small labels may become harder to read even if the new candidate meets the budget.

A 1200 by 900 image remains 1200 by 900 under a 1600-pixel maximum because this tool never upscales. A smaller source is not stretched to satisfy a nominal edge length. If the client requires a minimum width that the source lacks, obtain a suitable source rather than assuming this setting can create missing detail. If framing needs a crop, prepare it deliberately in the image resizer and identify the new composition clearly.

PNG needs a different expectation. The tool performs one lossless canvas encode and checks its size; it ignores quality and does not search progressively lower PNG settings. A PNG miss therefore cannot be solved by moving the quality control. You may consider a permitted dimension change, a different format or a revised source, but each changes the decision context. Preserve transparency and text requirements when assessing whether an alternative remains an acceptable asset.

Review and package a client-ready set

Inspect original and output side by side, then use the bounded 100 percent view on the details named in the brief. For the reception photo, read the entrance sign. For the screenshot, follow the route and read the street labels. For the banner, check every time and punctuation mark. A fitted overview is useful for composition but can hide the local defects that decide whether a small delivery copy is actually useful.

Before creating a batch, confirm the tool’s source bounds: up to 15 still JPG, PNG or WebP images, 20 MB each, 80 MB total, 25 megapixels and 8192 pixels per side. Animated or unsupported content is rejected rather than silently converted to an arbitrary frame. Processing runs locally and sequentially. These constraints govern preparation in this browser tool; they do not tell the client what a receiving platform will accept.

Use a selected-image download for an approved individual result or the ZIP for all encoded outputs plus the CSV summary and README. The ZIP is not an approval filter: target misses and enlarged outputs remain outputs. Check the numbered filenames against the source mapping, then assemble the client’s approved folder deliberately. If the directions screenshot remains unresolved, say so in the handoff instead of letting its presence in an archive imply approval.

Keep the CSV or copied summary when exact figures are useful to the recipient. The PDF provides settings, results and review thumbnails, not original images or a resumable session. Keep masters separately and label the chosen delivery revision. Changing settings or cancelling invalidates pending results and exports, so generate the handoff only after the current run is settled. Review files already downloaded from older runs separately; they do not update when you change a setting.

Resolve remaining requirements in the receiving workflow

The compressor’s destination selector is an intended-use label, not a native uploader or compatibility validator. Choosing a platform does not change the sampled quality sequence, supply automatic dimensions or convert a client allowance into an official posting rule. A technically valid image can still be the wrong asset for a particular placement. Use the social media specs checker for a separate preparation review when its supported checks match the task.

Document the final disposition of each client asset. In our example, the reception photo fits pending visual approval; the directions screenshot misses and needs an agreed change; the banner fits but offers no size saving over its source. That is a more useful handoff than a blanket batch-success message. Assign responsibility for the unresolved file and identify which condition the client has agreed to reconsider before another processing run begins.

Do not confuse local handling with archival preservation or privacy certification. Browser canvas processing may change color, profiles or metadata. Keep the authoritative source if those properties matter, and inspect brand-sensitive colors against the approved reference. The tool does not remove every possible sensitive detail or verify permission to share an image. A perfectly compliant byte count says nothing about a visible visitor name, location clue or confidential label left in the picture.

Record later delivery or review work in the content calendar generator when it belongs in a wider content plan, or explore the free tools directory for another preparation task. Images are not transferred into SMTasker and no native posting happens here. The completed job is a set of individually checked files with honest exceptions, plus a clear owner for whatever the receiving workflow still requires.

Questions about Compress Images to a Target Size

Continue with another image guide

Prepare your files in the bulk image compressor, or work through another image review below. More browser-based helpers are in the free tools directory.