Images and PDF

Resize images by pixels or percentage: fit, crop and export locally

·

To resize an image, choose its pixel dimensions or a percentage, decide whether to preserve its proportions or fill an exact box, then inspect the exported file. ToolMellow processes still JPG/JPEG, PNG and WebP images in your browser. Start with Fit inside when you want the entire picture without distortion.

The default is a 1200 × 1200 bounding box, Fit inside, enlargement off and JPG output at quality 90 with a white background. A wide photo may therefore export at 1200 × 800, rather than as a padded square. Choose PNG or WebP explicitly when transparent pixels matter.

Resize image

Choose the shape of the result first

A dimension requirement can mean either “stay within this box” or “produce exactly this width and height.” Fit inside answers the first request. Crop to fill and Stretch answer the second in different ways. Select the fit mode before judging the expected dimensions.

Consider a 400 × 200 source and a 300 × 300 box. Fit inside produces 300 × 150 without padding. Crop to fill produces a square by discarding edges; Stretch makes a square by changing proportions. Those last two examples require Allow enlargement because the requested height exceeds the source height.

Choose the shape of the result first
ModeOutput for the exampleWhat happens to the picture
Fit inside300 × 150; enlargement offPreserves proportions and all content; no added border
Crop to fill300 × 300; enlargement onPreserves proportions while removing edges
Stretch300 × 300; enlargement onKeeps all content but changes proportions

MDN: drawing source and destination image rectangles

A practical resizing workflow

Choose images, then wait for the previews to settle. Set Pixels or Percentage, choose an output format and process the batch. The source preview is a thumbnail; its displayed dimensions refer to the decoded original. A failed thumbnail is not an exported result: final processing still validates the file.

Download a result and check its actual dimensions, transparency and byte size in the application that will receive it. A smaller thumbnail or a lower quality slider value alone does not prove that an upload requirement has been met.

  • Keep an original copy before changing dimensions, crops or output format.
  • Use Fit inside for a complete photo; use Crop to fill only when an exact frame is required.
  • Choose PNG or WebP for alpha, and inspect a JPG background before downloading.

Pixels, presets and the requested box

Pixel mode accepts positive whole-number width and height values. The four presets set 1080 × 1080, 1200 × 630, 1080 × 1920 or 1920 × 1080. They are convenient sizes; check the destination’s current requirement rather than treating a preset as a guarantee for every platform.

In Fit inside, the entered values remain an upper box. The output canvas follows the fitted dimensions, so this mode does not create letterboxing or a fixed-size background. With enlargement off, a source already smaller than the box keeps its oriented dimensions, although it is still re-encoded.

Percentage applies to each image after rotation

Percentage mode scales each source independently from 1 through 400. A 400 × 200 image at 50% becomes 200 × 100. Selecting a 90° rotation first makes its working dimensions 200 × 400, so the same 50% exports 100 × 200. Each calculated side is rounded to a whole pixel, with a minimum of one.

The number control declares a whole-number step, while processing also accepts finite fractional percentages within the range. For example, 33.3% of 400 × 200 rounds to 133 × 67. Values above 100 require Allow enlargement; percentage is a geometric scale, not a file-size or visual-quality percentage.

Ratio lock links the target fields

Lock the current width/height ratio captures the ratio of the values you have entered. Starting at 90 × 30 and changing width to 60 links height to 20. It does not automatically recover every image’s original proportions or apply a different target ratio to each member of a batch.

Use first image dimensions fills the target fields from the first decoded, EXIF-oriented preview and updates the linked ratio. That preview size is taken before your explicit rotation setting. Selecting a quarter turn separately does not make the button swap the target width and height. Fit inside remains useful when a batch contains different shapes.

MDN: createImageBitmap orientation and color-space options

Position a crop without stretching the subject

Crop to fill keeps proportions and exports the exact target dimensions. Its horizontal and vertical position sliders run from 0 to 100, with 50 in the middle. Move toward an edge to retain that side when the source is wider or taller than the required frame.

A position control only affects an axis with surplus content to crop. It does not detect a face or select a subject automatically. For a manually chosen rectangle, use Image Cropper; the resizer’s controls describe the position of an automatic fill crop.

MDN: drawing source and destination image rectangles

Enlargement creates pixels, not recovered detail

Allow enlargement permits larger output geometry. With it off, Fit inside caps the scale at the source size; Crop to fill and Stretch reject a target when either side exceeds the corresponding oriented source side. A smaller width can still accompany a disallowed larger height.

Enlarging estimates additional pixels and cannot restore detail absent from the source. The browser draws with smoothing enabled and a high-quality hint; this is not a selectable interpolation algorithm, AI enhancement or a promise of identical results across browsers. Review fine text and edges at the intended viewing size.

MDN: image smoothing quality and browser supportAdobe: image size, resolution and resampling

Choose the output format and transparency

Image Resizer offers JPG, PNG and WebP output. Choose the format explicitly: this route does not have a Keep source format option. A browser that cannot return the requested encoding produces an error rather than a silently mislabeled result.

JPG flattens transparent pixels onto the chosen background, white by default. It does not remove an opaque background from a photo. PNG and WebP can keep alpha. Check a downloaded transparent result over both light and dark backgrounds, especially around softened edges.

Choose the output format and transparency
OutputTransparencyQuality control
JPGComposites alpha onto the chosen background10–100 encoder setting
PNGCan retain transparent pixelsSlider disabled; PNG encoding is lossless
WebPCan retain transparent pixels10–100 browser encoder setting

MDN: OffscreenCanvas output format, quality and resolution metadata

Dimensions and file size answer different questions

The quality slider is an encoder parameter for JPG and WebP, not a percentage of retained detail or a target number of kilobytes. Quality 100 does not guarantee pixel identity. PNG’s lossless encoding preserves its newly rendered pixel array, which may already differ because of resizing, cropping or format conversion.

This resizer always exports a newly encoded image, even at 100% or unchanged dimensions. A simple source can become larger after re-encoding. There is no exact-original fallback here and no target-KB control. Compare the reported source and output byte sizes; reduce dimensions or use Compress Image when byte size is the main constraint.

MDN: OffscreenCanvas output format, quality and resolution metadata

Orientation, rotation and flips

Supported source decoding applies EXIF orientation before calculating image dimensions. The explicit Rotate clockwise control then offers 0°, 90°, 180° or 270°. Horizontal and vertical flips refer to the output axes after that rotation.

These controls share the resize job’s settings across the batch. Check an asymmetric image after a turn and a flip instead of assuming that a horizontal flip always means the original camera’s horizontal axis. Changing a setting clears the old results, so process again before downloading.

MDN: createImageBitmap orientation and color-space optionsMDN: drawing source and destination image rectangles

Input, pixel and batch limits

Choose still JPG/JPEG, PNG or WebP files. GIF, SVG, HEIC, PDFs and remote image URLs are not inputs for this route. Supported headers and browser decoding determine acceptance, not merely a renamed extension. Detected animated PNG and WebP files are rejected rather than resized as animations.

Source headers, decoded dimensions and calculated outputs are bounded. The requested pixel box is checked before Fit inside reduces it, so a 6000 × 6000 request fails the pixel limit even for a small source. Reducing the target cannot bypass an oversized source. Allowed bounds still do not guarantee successful decoding on every device.

Input, pixel and batch limits
LimitValueScope
Files1–20One processing job
One input25000000 bytesDecimal 25 MB per file
Combined inputs100000000 bytesDecimal 100 MB per job
Encoded image results100000000 bytes combinedSeparate output bound; ZIP overhead is not included
Image geometry12000 pixels per edge and 24000000 pixels totalSource headers, decoded source, requested pixel box and calculated output
Percentage1–400Each oriented and explicitly rotated source
Worker deadline60 secondsWhole job, not a separate allowance for every file

Order files, download a ZIP or continue to PDF

Adding another selection appends to the queue. Move files up or down, sort by filename or reverse the order before processing. Output names combine a two-digit order prefix with a sanitized source base name and the chosen extension. Even identical source names receive different position prefixes.

Download each result separately or use Download all as ZIP for multiple results. The ZIP is packed locally on the browser’s main thread; it is not a further image-compression setting. Make a PDF or Make one PDF hands the selected image results to JPG to PDF once in the same tab. This handoff is held in memory and disappears on reload.

EXIF, print resolution and color

The canvas export does not copy the original EXIF metadata. The browser encoder can add its own metadata, so resizing is not a general privacy-redaction guarantee. Inspect the downloaded file if metadata removal is a requirement, and keep the source separately.

Pixel dimensions are different from print resolution and physical size. This tool has no DPI/PPI field or print-profile conversion workflow. Browser canvas encoders can supply their own resolution metadata; use a print-aware application when physical dimensions, embedded profiles or managed color are part of the specification.

MDN: createImageBitmap orientation and color-space optionsMDN: OffscreenCanvas output format, quality and resolution metadataAdobe: image size, resolution and resampling

Recover from a rejected or interrupted job

A batch is returned only after every file succeeds. If one file is invalid, the workspace does not offer a partial completed batch; remove or replace the problem file and try again. A format/header rejection, invalid geometry, enlargement restriction or device resource failure needs a different remedy.

Cancel stops the active worker and leaves the original selected files available. Try fewer or smaller files after a timeout. Clear files empties the queue and results, while the main resize settings remain. Processing image bytes happens locally, but visiting the site still loads the page and code over the network.

Recover from a rejected or interrupted job
SymptomLikely decision to checkNext action
Result smaller than the boxFit inside and enlargement offUse the actual dimensions, or choose an exact-box mode deliberately
Cannot enlarge messageA target side exceeds the oriented sourceReduce that side or enable enlargement
Valid-image messageUnsupported, animated, malformed or unreadable sourceRe-export a supported still image and retry
Timeout or local-engine failureBatch/resource or browser limitationCancel, reduce the job and process again

Check the downloaded result against its purpose

Inspect dimensions, crop edges, readable detail, transparency and byte size in the receiving application. An export that succeeds is ready for inspection; it is not proof of compliance with every upload, print or accessibility requirement.

  • For a complete photo, confirm that Fit inside retained the edges and did not distort proportions.
  • For an exact frame, inspect the crop position or stretching at the requested dimensions.
  • For a batch, confirm order, output filenames and every result before discarding any originals.

Common questions

Why is my resized image smaller than the chosen box?

Fit inside preserves proportions within a bounding box and does not add padding. A 400 × 200 source in a 300 × 300 box becomes 300 × 150. Enlargement off also prevents a smaller source from being scaled up.

How do I resize without stretching or cropping?

Choose Fit inside. It keeps the entire oriented image and its proportions, but the actual output may be smaller than one or both entered dimensions.

Can I enlarge an image without restoring missing detail?

Allow enlargement permits larger dimensions, but the browser estimates added pixels. It cannot recover detail missing from the source, and there is no AI enhancement control.

Does the percentage apply before or after rotation?

After the explicit rotation. A 400 × 200 source rotated 90° has working dimensions 200 × 400; 50% produces 100 × 200, with each side rounded to a positive whole pixel.

Why does the resized file have a larger byte size?

Every result is re-encoded, including at unchanged dimensions. Format, image content and encoder quality affect bytes. The resizer has no exact-original fallback or guaranteed target-KB setting.

How do I keep transparent pixels when resizing?

Choose PNG or WebP output and inspect the alpha edges. JPG composites transparent pixels onto the chosen background. It does not remove an already opaque background.

What are the image and batch limits?

Each job accepts 1–20 still images, up to 25000000 bytes each and 100000000 bytes combined. Encoded results have a separate 100000000-byte total bound. Geometry is limited to 12000 pixels per edge and 24000000 pixels overall; the requested pixel box is checked too.

Can I download resized images together as a ZIP?

Yes. Multiple results offer Download all as ZIP, named toolmellow-images.zip. Packing occurs locally on the main browser thread. Download individual results if ZIP creation fails.

Will resizing keep the original EXIF and print resolution?

Original EXIF is not copied through canvas export. Browser-added metadata can remain, and there is no print-resolution-only control. Check metadata and use a print-aware application for physical size or managed profiles.

Does ratio lock use each source image’s proportions?

No. It captures the current target width/height ratio and links those two fields. Use first image dimensions takes the first EXIF-oriented preview size before explicit rotation; shared batch settings still apply.

Can I resize animated images, SVG or HEIC here?

This route accepts supported still JPG/JPEG, PNG and WebP images. Detected PNG/WebP animation is rejected; SVG, GIF and HEIC are not supported inputs. Use HEIC to JPG for an appropriate HEIC conversion first.

Does quality 100 make resizing lossless?

No. JPG/WebP quality is a browser encoder parameter, and resizing or cropping changes pixels. PNG encoding is lossless for the newly rendered pixels; it does not undo a geometric change.

What happens when one file in a batch fails?

The job returns an error without a partial result list. Remove or replace the rejected file and process again. Changes to the queue or settings clear stale results.

Are my images uploaded, and can I cancel?

Image bytes are processed on your device. Cancel terminates the active worker and retains the originals in the queue. The initial page/code load uses the network; the same-tab PDF handoff is temporary memory, not persistent storage.

Primary sources and technical references

Put it into practice.

Related guides