Image tools

Image work should begin with the place the file will be used. A product thumbnail, social card, transparent logo, email photograph, and sampled brand color have different requirements. Record the target pixel dimensions, aspect ratio, accepted formats, transparency need, file-size ceiling, and visual tolerance before processing. Resize changes pixel dimensions, Crop changes framing, conversion chooses an output encoding, Compress adjusts quality for supported lossy output, Blur changes appearance, Remove Image Metadata re-encodes pixels, and Color Picker reports one sampled pixel.

A small file is not automatically suitable, and a large pixel count is not automatically better. Upload systems can reject the wrong format even when size is low. A correctly sized image can crop the subject badly. PNG can preserve transparency but be unnecessarily heavy for a photograph. JPEG can be compact while replacing transparent areas with white in this workbench. WebP can reduce delivery weight but still needs destination support. Make each choice against the final requirement, then verify the exported image there.

Resize Image starts with the source width and height. Changing width recalculates height from the original aspect ratio, and changing height recalculates width in the same way. The canvas output uses the resulting integer dimensions, with each side constrained to at least one pixel during processing. This suits proportional scaling for a known target width or height. It does not expose an unlock control for intentional stretching, a resampling method, resolution metadata, batch sizes, or responsive image variants.

Scaling down removes pixel detail that cannot be recovered by enlarging the exported copy later. Scaling up creates more pixels through browser canvas interpolation but does not reveal new source detail. Keep the original and generate delivery sizes from it rather than repeatedly resizing derivatives. Check sharp edges, small text, faces, and line art at actual display size. Pixel dimensions alone do not describe print dimensions because print workflows also depend on physical size and resolution assumptions outside this interface.

Crop Image uses numeric x and y coordinates plus width and height, all measured in source pixels. The rectangle must stay inside the image, and width and height must be positive. The output canvas matches the crop dimensions exactly. This is a precise rectangular crop rather than an interactive drag overlay. Calculate the intended boundaries from the source dimensions, then preview the result. The tool does not provide preset social ratios, face detection, horizon correction, rotation, or automatic subject positioning.

Cropping changes what the image says. Removing surrounding context can improve focus, but it can also exclude a product part, attribution, scale reference, person, disclosure, or safety information. Review every destination crop instead of assuming one framing works across cards, banners, and thumbnails. Keep important content away from edges because downstream platforms may crop again. If you need several aspect ratios, return to the original for each crop rather than trimming an already narrowed derivative.

Coordinate errors are easiest to catch with known boundaries. The top-left corner is x zero, y zero. Width extends right and height extends down. A rectangle whose x plus width exceeds the source width is rejected, as is one whose y plus height exceeds source height. If a crop fails, compare all four values with the detected dimensions. Avoid guessing by repeatedly reducing values when exact framing matters; derive the rectangle and document it for reproducible assets.

JPG suits photographs when transparency is unnecessary and lossy compression is acceptable. PNG keeps an alpha channel and sharp raster edges but can make photographs large. WebP and AVIF can reduce delivery size, although destination support and encoding time vary. HEIC is useful in Apple photo workflows but is less common on the web. Every rewrite applies orientation, converts decoded pixels to 8-bit sRGB, and removes source metadata.

GIF has a limited color palette and binary transparency. BMP favors compatibility with older software at the cost of large files and no alpha in this converter. TIFF supports alpha and multiple image directories for scanning, print, and archival workflows. ICO is limited to one image no larger than 256 by 256 pixels. SVG is accepted only as a source and is rasterized after active content, embedded images, and external references are rejected. Raster files are never presented as vector SVG output.

Animated WebP, GIF, and multi-page TIFF keep every frame only when the destination supports multiple frames. Other destinations export the first frame and identify the result as a still image. JPG and BMP require a selected background when decoded pixels contain transparency. Re-encoding can change pixels, color representation, bit depth, metadata, transparency precision, frame behavior, and file size, so no format route is described as lossless.

Compress Image exposes a quality slider from 10 to 100, starting at 82. Output type follows the source among PNG, WebP, and otherwise JPEG. Browser canvas encoders do not necessarily use the quality number identically across formats, and PNG may not respond as users expect from a lossy-quality control. The ready message reports output size as a percentage of the source. That number can exceed 100. Compression should therefore be judged from both downloaded byte size and visible quality.

Create a small comparison set rather than lowering quality until a file merely passes a limit. Inspect the result at 100 percent zoom and at its real display size. Watch faces, hair, foliage, text, icons, gradients, and flat color boundaries. A slightly larger file may avoid distracting artifacts, while a smaller dimension can sometimes save more bytes than harsh quality loss. This workbench does not target an exact kilobyte value, compare several outputs side by side, or calculate perceptual quality.

Repeated lossy saves accumulate damage. Return to the highest-quality source for each experiment and keep the approved setting with the asset record. File-size comparisons can also be misleading when the source contains metadata or uses a different encoder. A large reduction may partly come from metadata removal rather than pixel encoding, while a rewritten small source can grow. Verify loading performance in the actual page and ensure responsive delivery uses appropriate dimensions instead of sending one oversized file everywhere.

Remove Image Metadata draws the decoded image onto a canvas and creates a new file. This normally omits many source metadata blocks because the canvas output is built from pixels rather than copied container data. Output type follows PNG, WebP, or otherwise JPEG. The page does not list discovered EXIF, IPTC, XMP, GPS, thumbnail, profile, or application fields before removal, and it does not verify the downloaded binary for every metadata type afterward.

Sensitive information can remain visible in pixels even when metadata is gone. A badge, address, reflection, face, screen, document, or location landmark requires visual review and possibly cropping or a proper redaction workflow. Blur Image applies one blur level to the entire image, not selected regions, and a blur may not be suitable redaction. Removing metadata also can discard useful color, orientation, copyright, or accessibility workflow information. Keep the source and inspect the result with a dedicated metadata tool when policy requires proof.

Blur Image applies a browser canvas filter to the whole image, with a slider from zero to 40 pixels. It does not select a face, license plate, background, or rectangular region. Blur spreads pixels and can interact with canvas boundaries, format, and later resizing. Use it for a deliberate full-image effect, then inspect edges and subject recognition. Do not treat an attractive blur as irreversible redaction; source information may remain inferable, and the unblurred original must still be protected.

Color Picker draws a preview canvas scaled down only when the source width exceeds 900 pixels, then reports HEX and RGB values for the clicked preview pixel. Scaling can blend neighboring source pixels, so a sampled edge may represent an interpolated color rather than one original pixel. It reports red, green, and blue without an alpha value and does not name colors, build palettes, measure contrast, or account for full color-management workflows. Sample a flat region several times and validate critical brand values from the source specification.

JPG and PNG sources can use browser canvas for reliable JPG, PNG, and WebP output. Other registered pairs use the private ARM64 worker and its verified ImageMagick delegates. Input is limited to 25 MB, each side to 16,384 pixels, total decoded output to 50 megapixels, and animated input to 100 frames. Worker jobs also have fixed memory, disk, concurrency, output-size, and time limits. A small compressed file can still be rejected if decoding would exceed those bounds.

Browser jobs keep source pixels on the device. Worker jobs are validated by the application, validated again by the worker, processed in a unique temporary directory, and removed after success or failure. Neither path stores image content in the database or analytics. The workbench has no persistent project, batch queue, undo history, layer model, or cloud recovery. Keep the source elsewhere and download each approved result before resetting or leaving the page.

After processing, confirm filename, format, byte size, pixel dimensions, aspect ratio, transparency, crop, orientation, and visible quality. Zoom into detailed regions, then view the image at its expected rendered size. Open it in another application and upload it to a safe test destination. Check that a transparent image remains transparent, a JPEG background is intentionally white, and a WebP is accepted. Compare color-critical assets on the target display or production workflow rather than trusting one browser preview.

For web publishing, inspect responsive layout, alternative text, intrinsic width and height, loading behavior, and any further server-side transformations. For print, obtain a proof at final physical size and confirm color and sharpness with the vendor. For privacy, review both pixels and metadata under the applicable policy. For a color sample, compare several nearby points. A ready status confirms that canvas produced a file; approval requires evidence from the context that will deliver or display it.