Find the images making your PDF large, then change only those

Read the image list before choosing a setting

The list reports each indirect image object's pixel dimensions, stored stream bytes, encoding, color space and page-resource references. Those image bytes are counted once per object. They exclude the rest of the PDF's structure, fonts, text streams and other data, so their sum is not a complete file-size breakdown.

PDF page drawing instructions refer to resources by name. The same image object can therefore be referenced in more than one place. This tool follows page resources and nested Form resources; it does not count how often an image is actually painted. Resource lists can include unused entries, and inline images are not listed.

Reference: Adobe / ISO 32000-1: resource dictionaries, image XObjects and document outlines

Try a selective optimization

1. Choose an unsigned, unencrypted PDF. Start with the largest eligible image payloads and check their listed original pages.
2. Select the image objects to change. Set JPEG quality and Maximum image dimension (px). The dimension limit never enlarges an image, and its placement on the page remains unchanged.
3. Choose Preview selected image optimization. The worker re-encodes only the selected supported images; it does not turn text or vector page instructions into screenshots.
4. Compare the before/after image previews and the original/result PDF byte counts. Review the PDF preview, then download and inspect fine detail in a local reader.
5. Choose Download image-optimized PDF to keep all original pages, or continue below to select, reorder, rotate or split pages. Use original images instead discards the image changes.

The small before/after image previews have a 360-pixel longest side. They help identify the changed content but cannot establish that small lettering, diagrams or photographic detail survived acceptably. Inspect the exported PDF at the size at which people will read or print it.

Use measured file bytes to judge the result

JPEG quality is a lossy encoder setting, not a percentage of retained visual information. Browser encoding can also change colors. A lower setting may introduce artifacts around text and edges; compare the result rather than treating the number as a quality score.

The tool keeps an individual source image unchanged when its new JPEG payload would be at least as large. Even when selected image payloads shrink, serializing the PDF can change its structural overhead. The whole-file before/after count is the relevant measurement for an upload limit; savings are not guaranteed.

Reference: MDN: OffscreenCanvas JPEG format and quality options; pdf-lib PDFDocument: page copying and serialization

Review bookmarks when removing or reordering pages

Bookmarks are navigation entries with destinations; they are not the page content itself. Retained supported destinations follow their original page through reordering. If you remove a bookmarked page, the organizer requires a decision: delete the entry or retarget it to a selected original page. Unsupported destinations or actions also require review.

Retain keeps supported titles and hierarchy, while opening views reset to Fit page. Rebuilt bookmarks never copy external or unsupported actions. Deleting an entry keeps retained descendants under the nearest retained ancestor. A split export includes destinations for its own page and relevant parent headings. Check navigation in the downloaded output.

Choose Create PDF for the selected pages in their displayed order, or enable Split selected pages into separate PDFs (ZIP). Ordinary form fields are flattened during page organization. Document attachments and document-level metadata are not preserved by that page-copy workflow, so keep the source.

Reference: Adobe / ISO 32000-1: resource dictionaries, image XObjects and document outlines; pdf-lib PDFDocument: page copying and serialization

Check the supported image types and document limits

Image optimization supports 8-bit DeviceRGB and DeviceGray images encoded as a single JPEG filter, raw samples or supported zlib/Flate samples, including PNG predictors. Masks, explicit decode arrays, external or alternate images, ICC/indexed/CMYK color spaces, filter chains and other codecs stay unchanged. This deliberately limited conversion is not a color-managed production-print workflow.

Upload limits are 25 MiB and 200 pages, with a 50 MiB total export budget and 15-second processing jobs. The image list allows up to 500 objects; each supported image is limited to 16 megapixels and 16,384 pixels per side. Select up to 25 images and 40 megapixels per optimization. JPEG quality runs from 20 to 95, with a maximum dimension setting from 256 to 8,192 pixels.

Signed, certified, signature-field, encrypted and XFA PDFs are rejected. Bookmark review is bounded to 500 entries, 12 nesting levels and 500 characters per title. A split also requires source bytes multiplied by selected pages to fit the export budget. Worker image decoding and JPEG encoding must be available in your browser; page organization remains available when the image encoder is unavailable.

Reference: MDN: OffscreenCanvas JPEG format and quality options

Technical references

These primary sources explain the underlying formats and browser APIs. Tool-specific behavior and limits above describe the current Buddy Code implementation.

• Adobe / ISO 32000-1: resource dictionaries, image XObjects and document outlines
• pdf-lib PDFDocument: page copying and serialization
• MDN: OffscreenCanvas JPEG format and quality options