Skip to content

Image tools: crop, compress, convert, and enlarge

Open Utilities → Image tools for Crop, Compress, Convert, and Enlarge. Keep the original, then check text, dimensions, backgrounds, and clarity in the downloaded file. The original workflow diagrams and format guidance were checked against real file selection and downloads on October 10, 2026. The rotation, mirror and previous-download example below was added from the October 13, 2026 native checks.

The web and desktop apps open tool pages that must finish loading before use. These are separate from Magic Pic's image translation, editing, AI generation, and AI enhancement. For perspective correction, shadow removal, handwriting cleanup, or combining page photos into a PDF, use the scan workspace.

Choose a tool and prepare inputs

ToolInputMain output
Crop (裁剪)Browser-readable images; select several and process them individuallyJPEG, PNG, WebP, or a copied result image
Compress (压缩)One or more imagesJPEG, WebP, AVIF, or optimized PNG; individual files or ZIP
Convert (转换)One or more imagesJPEG, PNG, WebP, AVIF; individual files or ZIP
Enlargement (无损放大)One readable PNG, JPG or WebPAn image enlarged to 2× or 4× dimensions

The picker accepting image files does not guarantee support for every camera or specialist format. If a file cannot be read, use an image application to save a genuine PNG/JPEG copy rather than merely renaming its extension. Large images and batches depend on available device memory; reduce dimensions or batch size if needed.

On narrow phone screens, settings are often behind the purple menu button. Open it to find options and processing/download controls. Adding a file to a tool does not overwrite the original on disk.

Crop: keep the required region

  1. Open Crop, then click or drag to add an image. When several are present, select the current one and retain the original.
  2. Choose Free crop or a ratio such as 1:1, 4:3, 16:9, 3:2, or 9:16, output format and quality before choosing orientation. The left/right buttons turn by 90°; Horizontal flip also reverses text.
  3. Set Grid snapping and EXIF orientation correction as needed. The EXIF option concerns camera orientation metadata and was not tested by the direction sample below. Recheck the selection after changing orientation, format, ratio or other options.
  4. Adjust the frame and corners last, retaining complete questions, legends, page numbers and required edges. If a rotated image extends beyond the view, reduce the selection first, zoom out with the wheel to see the image, then expand its corners to the required bounds.
  5. Apply Crop image to generate this result. Generate again after changing settings, then download; an existing download control may still refer to the previous output.
  6. Inspect the new result and its actual dimensions, then download it or copy it to an app that accepts images. Check the settings and selection again for each subsequent image.

Explicitly choose JPEG, PNG, or WebP when the format matters. The current Original (原样) option encodes JPEG, so it does not guarantee the original encoding or transparency. The crop canvas uses a white background; inspect transparent images after downloading. The current page has no custom width/height input. Dimensions follow the crop region and canvas limits; check the downloaded dimensions.

Choose the crop for the selected image and check the downloaded result. (workflow diagram)

Choose the crop for the selected image and check the downloaded result. (workflow diagram). Open the image in a new tab to enlarge it.

Rotation, mirroring and an earlier download: one real example

On 2026-10-13, a self-authored, static, opaque 640×400 PNG was processed in the Chinese image-crop workspace. Its corners are marked A/red, B/green, C/blue and D/purple, with an upward arrow and normally oriented UP 123 text. This example used PNG, free crop, grid snapping off and EXIF correction off.

Actual outputPixel width × heightTop left / top rightBottom left / bottom right
First 90° output without expanding the default selection1080×675D / AB and C are missing
90° with full selection675×1080D / AC / B
180° with full selection1080×675C / DB / A
270° with full selection675×1080B / CA / D
0° with horizontal mirroring1080×675B / AC / D

The first actual output rotated by 90° uses the default selection and omits the green B and blue C corner markers.

First actual download: a completed rotation does not establish that the full image was retained.

The actual PNG rotated by 90° after selecting the whole image retains all four corners; the arrow points right and output dimensions are 675×1080.

After zooming out and expanding the selection to all four edges, a new crop retained every marker. The source was 640×400 and this result is 675×1080, so it was resampled; unchanged source dimensions or lossless rotation are not established.

Changing settings did not change the existing result. After the 90° output, the settings were changed to 0° plus horizontal mirroring. Downloading before another crop returned exactly the previous 90° bytes. Applying the crop again produced the 1080×675 mirror, including reversed text. Confirm orientation and format, then check the final selection and generate the result.

Download the source image, full 90° PNG, 180° PNG, 270° PNG, horizontal mirror, or the complete ZIP with the cropped first attempt, repeated old download and checks.

The five generated outputs and one extra old download came from real operations; output bytes are unchanged and the ZIP was assembled locally. This example did not test EXIF camera orientation, transparent input, animation, arbitrary angles, vertical flipping, batch application or PDF rotation saving. The image-direction guide provides the complete comparison in Chinese.

Compress: reduce file size

  1. Open Compress and add the sample images. Add more in the workspace if needed; split the batch when the interface reaches its limit.
  2. Choose a Quality level: high, standard, medium or low. The current workspace has no target-KB input, compression-mode switch or custom quality slider.
  3. Choose JPEG, WebP, AVIF, or PNG optimization. Try JPEG/WebP for photos; for transparency, choose a supporting format and verify the result. AVIF requires a compatible browser encoder; choose another format if encoding is unavailable.
  4. Check the quality and format for this run. On a phone, close the settings panel and return to the file list to start compression. Changing a setting clears earlier results, so generate them again.
  5. Start compression and wait for individual results. Compare original size, compressed size, and previews; zoom in on small text and thin lines.
  6. Download successful images individually or download all as ZIP. Duplicate names receive suffixes. Extract the archive and inspect file count and contents before removing any originals.

For a limit such as 50KB, 100KB or 200KB, generate a candidate with a quality preset and inspect its actual byte count. The tool preserves pixel dimensions and does not guarantee an arbitrary target size or fixed reduction. If text becomes blurry, increase quality and start from the original. If the output is larger, keep the original or compare another allowed format.

Compare file size and small-text clarity before downloading. (workflow diagram)

Compare file size and small-text clarity before downloading. (workflow diagram). Open the image in a new tab to enlarge it.

Same-name image batches: ZIP identity, size, and transparency

A batch shares settings, but each image still needs its own size, background, and destination checks. On October 13, 2026, we copied two previously published, self-authored PNGs into separate local folders and named both figure.png, then selected them together. The chart is 1600 × 1000, 101,179 bytes; the transparent shapes image is 800 × 500, 20,417 bytes.

  1. Confirm File list (2). Use original size and content to distinguish inputs.
  2. Select Standard (75%) → JPEG, run Start compression, and use Download all after completion.
  3. Selecting Low quality (45%) cleared old results and disabled Download all. Run compression again before downloading the new JPEG ZIP.
  4. With the same original inputs, select Standard (75%) → WebP, process again, and download. We also downloaded the second WebP individually from its file row.
  5. Extract each ZIP and check both files, exact bytes, dimensions, and content. One passing file does not establish that the whole batch meets the limit.

Actual batch workspace with two files named figure.png; standard WebP results are displayed as 38.82 KB and 9.86 KB.

Output in this checkChart: 1600 × 1000Transparent shapes: 800 × 500
JPEG, standard 75%69,414 bytes12,177 bytes; transparent margins become white
JPEG, low quality 45%48,645 bytes8,948 bytes; transparent margins become white
WebP, standard 75%39,752 bytes10,094 bytes; input alpha is preserved

Using 50,000 bytes as an example limit, standard JPEG passes only the second image. Both pass in the other two groups, but this does not guarantee an arbitrary batch will fit 50 KB. Check the recipient's byte convention, formats, and required image detail. The transparent original is already 20,417 bytes and may meet the need without re-encoding.

Each native ZIP contains two distinct images without overwriting. The WebP ZIP names the chart compressed_figure.webp and the second image compressed_figure (2).webp. The separate download of the second image is nevertheless called compressed_figure.webp, with bytes exactly matching the ZIP's second member. Suffixes distinguish entries inside a ZIP; they do not replace checking source identity.

Local comparison of actual JPEG and WebP outputs: JPEG has white margins, while WebP retains transparent margins and a translucent square. Checkerboard and labels are not in the downloaded images.

JPEG has no alpha channel; this example has opaque white margins. The WebP alpha channel matches the original PNG byte-for-byte, including 262,160 fully transparent pixels. That establishes this image's transparency retention, not lossless color: color pixels are compressed. All six outputs retain the original dimensions; 45% quality does not mean 45% of the width and height.

Download the same-name inputs, three native ZIPs, single WebP, and check notes. The outer bundle and input-folder ZIP were assembled locally. Native tool ZIPs and the single WebP remain unchanged. The comparison only arranges outputs for display. Each run used the original images, not an earlier low-quality output.

These are two static, self-authored graphics, not camera photographs. The interface lists a maximum of 20 images, but this check does not test that maximum, recursive folder import, animation, or EXIF. Use smaller groups for large images; preserve successful files and handle failures separately. Even after size checks pass, inspect small text, edges, and suitability for the destination.

Convert: change image format

  1. Open Convert, upload one or more images, and check the file list.
  2. Select JPEG, PNG, WebP, or AVIF as the output. SVG output is not currently available.
  3. Choose Keep EXIF as needed, then inspect metadata and colors in the result. There is currently no active sRGB color-profile conversion control.
  4. Start conversion and wait for each file's result or error. Use the completed, downloadable file rows to confirm success.
  5. Download individual files or a ZIP. Duplicate output names receive suffixes to retain every successful file. Check file count and extensions, then open the results in the receiving application.

JPEG composites transparent areas onto white; PNG and WebP can retain transparency already present in the input. Converting JPG to PNG leaves its existing white background in place and does not remove it. AVIF encoding depends on the browser; unsupported encoding shows a failure reason without a download disguised as another format. The receiving application must also support the format. This flow covers static images and does not guarantee animation frames or timing.

If one file fails, preserve successful results and retry the failed image on its own. If processing components fail to initialize, wait for loading or reopen the tool; download completed files before reopening.

Select the target format and check each converted image. (workflow diagram)

Select the target format and check each converted image. (workflow diagram). Open the image in a new tab to enlarge it.

Enlarge: increase dimensions to 2× or 4×

The Chinese menu uses the legacy label Lossless enlargement (无损放大); the current tool performs ordinary image resizing. Increasing the pixel dimensions cannot guarantee recovery of missing text or detail. Magic Pic's separate Image enhancement (图片高清化) provides AI processing with its own terms and result checks.

  1. Open Enlarge and choose one readable PNG, JPG or WebP.
  2. Select 2× or 4×. An 800 × 600 image becomes 1600 × 1200 at 2×.
  3. Start enlargement, wait for the result, and compare text edges and diagram lines with the original.
  4. Download the image and open it in your presentation or layout application to confirm the dimensions. Generate a new result after changing the scale; an earlier output is not offered as the new scale's download.

The current page has no sharpening, denoising or quality slider. Both language pages perform real image resizing; verify the dimensions of the downloaded file. For larger images, begin with 2× to reduce memory demands. If decoding fails, use an image application to save a PNG copy before retrying.

2× and 4× increase image dimensions; inspect the enlarged detail. (workflow diagram)

2× and 4× increase image dimensions; inspect the enlarged detail. (workflow diagram). Open the image in a new tab to enlarge it.

Downloads and recovery

Browser downloads appear in the download list or the location you select. In the desktop app, follow the actual download/save prompt. Previews and queues are temporary work areas; download required results before refreshing, closing, or replacing files.

ProblemRecovery
Image cannot be readSave a genuine PNG/JPEG copy in an image app, then retry; check that the source is not damaged
Settings changed but the download is the previous imageConfirm the final selection and apply the crop again, then inspect the new download
Copying or pasting failsDownload the image and insert the file instead
No result appearsWait for individual file completion, inspect errors, then reduce batch size or dimensions
Transparency becomes whiteCropping fills the background white; preserve the original and check PNG/WebP conversion results when transparency is needed
AVIF encoding is unavailableChoose JPEG, PNG or WebP and generate a new result; renaming the extension is not conversion
JPG converted to PNG still has a white backgroundThe white pixels are part of the JPG; obtain an original transparent asset or edit the background separately
ZIP cannot be inserted as an imageExtract it first, then select an image file inside
Tool or settings fail to openCheck connectivity and reopen the tool; on a narrow screen, use the purple menu button