creatorvalet

Resize and crop images in bulk

Apply one exact pixel rule, review only the risky crops, and receive a byte-checked ZIP without uploading the image batch.

  • SECURE
  • NO UPLOADS
  • NO SIGNUP
  • BROWSER BASED
  • FREE
  • FOREVER.

Support us with a link or a share

Choose images for the batch

Drop, choose or paste PNG, JPEG, WebP, GIF, AVIF or BMP files. Nothing is uploaded.

Resize and crop images in bulk with one visible rule

Add several PNG, JPEG, WebP, GIF, AVIF or BMP files by choosing, dropping or pasting them. Before any output is encoded, each card shows the format read from the bytes, available dimensions, likely transparency and animation. Select the images that belong in this batch and enter whole-pixel width and height.

The workbench is for a real repeated job, not a row of unrelated shortcuts. Its batch rule includes placement, upscaling, format, quality, padding and a filename pattern. A change invalidates the previous package so a download can never quietly retain settings from an older run. If the pixels should stay byte-identical and only the names need to change, use the bulk file renamer instead.

Bulk image crop and fit are different promises

Fill and crop scales an image until the target rectangle is covered, then removes what lies outside. The aspect ratio is preserved and the pixels are never stretched. A wide source entering a square target therefore loses its left and right edges. The card reports material loss before processing and gives horizontal and vertical focus sliders for an individual decision.

Fit with padding keeps the entire source. The unused target area is transparent, white or black according to the visible setting. If the output is JPEG, transparency is not possible and white is used instead. For logos, product shots and diagrams, fit is usually the safer first pass; for profile tiles and thumbnails, fill is usually the consistent one.

Small files are not enlarged silently

Allow upscaling is off by default. A 90 × 90 source headed for 200 × 200 stays 90 × 90 and is centered with padding. Turning the option on permits enlargement, but no resampling method can recreate detail the source did not contain. The resulting file has the requested canvas dimensions either way; the distinction is whether the original pixels were enlarged inside it.

Alpha is also a review signal. A transparent logo can look correct on the checkerboard but fail over a dark site, while JPEG must flatten it. Animated GIF and WebP files are flagged because this still-image workflow uses the first decoded frame. Use a purpose-built animation converter when motion is part of the material.

Review exceptions instead of recropping the whole batch

The same centered rule is applied to every selected image first. Cards with large crop loss, alpha, animation, insufficient source pixels or an unsupported decoder are marked for review. Open an individual crop only where its subject needs a different focus. This keeps a hundred-image job a batch operation without pretending that one center point is correct for every photograph.

If the images live inside Word, begin with extract images from a Word document. That route preserves byte-exact DOCX originals and hands the selected embedded images to this same workbench without asking for another upload. If the goal is fewer bytes rather than uniform dimensions, the image compressor owns that separate decision.

Every successful output and ZIP entry is read back

Encoding happens in a browser worker so a batch does not monopolize the page thread. Progress counts completed images; it does not advance on a timer. The generated PNG, JPEG or WebP is inspected again and must report the exact chosen width and height before it is accepted. Failures remain named in the manifest while successful images stay available.

The worker decodes one image at a time so a batch never multiplies raster allocations through concurrency. CreatorValet does not guess a pixel or file-size ceiling: it attempts the real browser decode and canvas allocation, then keeps a named per-image failure in the manifest if that platform cannot complete it. A single package is structurally limited to 20,000 selected images.

The final ZIP contains the dimension folder and CSV and JSON manifests. It is reopened in this tab, all entries are read, and their hashes are compared with the bytes the workbench intended to write. Nothing is posted to a resize API. The verification guide gives a repeatable browser check for that claim.

Questions

Can I resize multiple images to an exact size?

Yes. Width and height are unrestricted positive whole-pixel inputs rather than a short preset list. The real browser canvas is attempted, and every successful file is read back before download and must report exactly the requested dimensions.

What is the difference between fill and crop and fit with padding?

Fill and crop covers the complete target rectangle without stretching, so source pixels may fall outside it. Fit with padding keeps the whole source and places the unused area around it. Transparent, white or black padding is explicit.

Will small images be enlarged?

Not unless you turn on Allow upscaling. With the default off, a small source stays at its decoded pixel size and the remainder is padding. The card is flagged before processing so the result is not a quiet surprise.

What happens to animated GIFs?

They are flagged before processing. The resized output uses the first decoded frame; the tool does not pretend that a still-image canvas preserves animation. Keep the original GIF if motion matters.

Are my images uploaded?

No. Image bytes are decoded and encoded in a browser worker, then the ZIP is built and reopened in this tab. The selected images are never posted to CreatorValet or a third-party image service.