I have a 72.9MB PDF file that I need to shrink into under 500KB.

The file was a JPEG image that I had scanned, and then converted to pdf.

  • 1
    it depends on what consumes the space...need a lot more information. compressing image space could help, but if you're trying a large file heap spray, that won't work. seriously need more info. – RobotHumans Mar 16 '12 at 17:14
  • 1
    convert it to DjVu, instead trying to reduce to impossible sized PDF (according source) – zetah Mar 16 '12 at 17:22
  • the file was a jpeg image that I had scanned, and then converted to pdf. – tamimym Mar 16 '12 at 19:53
  • It only seems to help filesize a little bit, but pdfopt has a simple syntax and improves loading and page-turning speed in the iPad era. :-) – Ari B. Friedman May 31 '12 at 0:53
  • 1
    PDF to PS is not effective in scanned PDF file, I try to convert 56 MB pdf into ps file but ps file convert into 1.3 GB and again ps2pdf is converted in 45 MB file – user124118 Jan 18 '13 at 5:32

20 Answers 20

up vote 115 down vote accepted

aking1012 is right. With more information regarding possible embedded images, hyperlinks etc.. it would be much more easier to answer this question!

Here are a couple of script and command-line solutions. Use as you see fit.

  • 10
    Thank you very much for your suggestions, the ghostscript shell worked wonders and shrank it down to 460KB :) – tamimym Mar 16 '12 at 19:56
  • That is not necessarily true. If the content went from an image to text, that is more than fessible. [That is assuming that the text is accurately ocred] – monksy May 16 '15 at 20:11
  • 2
    I recommend you shrinkpdf.sh script, you can customize the code to use the ppi value you want (72 by default) and reach exactly the filesize you need to sacrifice the least quality. This made me able to upload a scanned document of 11 MB with a max. size of 3 MB without losing a lot of quality. – Severo Raz Apr 9 '16 at 22:18
  • 4
    shrinkpdf works great! – AmanicA Feb 22 '17 at 22:14
  • First link, with setting "/ebook", reduced a 19MB scanned file to 4.2MB and the scanned text remains readable. – dremodaris Dec 13 '17 at 21:10

Use the following ghostscript command:

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/screen -dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf input.pdf
  • -dPDFSETTINGS=/screen lower quality, smaller size.
  • -dPDFSETTINGS=/ebook for better quality, but slightly larger pdfs.
  • -dPDFSETTINGS=/prepress output similar to Acrobat Distiller "Prepress Optimized" setting
  • -dPDFSETTINGS=/printer selects output similar to the Acrobat Distiller "Print Optimized" setting
  • -dPDFSETTINGS=/default selects output intended to be useful across a wide variety of uses, possibly at the expense of a larger output file
  • 2
    One can also make a Nautilus script to access this function for every file. – Sina May 7 '13 at 15:51
  • 3
    This should be the accepted answer. ghostscript is the PDF, XPS and PS implementation for unices and can do basically everything delivering best quality... – dom0 Oct 2 '13 at 17:27
  • 7
    @Sina: There is actually a Nautilus Script with a simple Zenity-based GUI that utilizes this gs command with all its quality-level options: launchpad.net/compress-pdf – Sadi Oct 25 '13 at 11:47
  • 25
    This is the right answer for this question (compressing a pdf that is mostly bitmap data). I found that the screen setting was too low quality for me, but ebook worked well, cutting a 33Mb scan-based PDF down to 3.6Mb, and keeping it very readable. Other options for the -dPDFSETTINGS option are listed here: milan.kupcevic.net/ghostscript-ps-pdf, and it might be a good idea to include them in this answer. – naught101 Dec 2 '14 at 1:13
  • 4
    gs available configuration parameters: ghostscript.com/doc/current/Ps2pdf.htm – Antonios Hadjigeorgalis Dec 11 '14 at 23:41

My favorite way to do this is to convert the pdf to ps and back. It does not always work, though, but when it works the results are nice:

ps2pdf input.pdf output.pdf

This also directly works on pdf's, as suggested in the comments.

  • 3
    This is a very simple and effective way to do it. I was surprised to see how much this method compressed the files. Thanks you! – Gabriel Apr 8 '13 at 12:59
  • 14
    Despite the fact that this one approach became my favorite solution to compress pdf files, it breaks up url links the document may have (which does not happen with @Michael D's approach). Apart from that, awesomeness is all I can think of running this snippet! (: – Rubens Dec 6 '13 at 11:01
  • 1
    @Rubens Ah. Did not know about the fact that it breaks the url links. Thanks for adding that. – don.joey Dec 6 '13 at 12:19
  • 2
    This bypasses password protection...just sayin' – jojo Jan 6 '15 at 20:28
  • 4
    ps2pdf will take pdfs as inputs, so you can do this in one step: ps2pdf intput.pdf output.pdf – frabjous Sep 1 '16 at 19:19

If you have a pdf with scanned images, you can use convert to create a pdf with jpeg compression (You can use this method on any pdf, but you'll loose all text informations).

For example:

convert -density 200x200 -quality 60 -compress jpeg input.pdf output.pdf

Adjust density (e.g. 100x100) and quality to your needs.

Depending on your input jpeg compression might not be the best choice due to compression artifacts. You have the choice between BZip, Fax, Group4, JPEG, JPEG2000, Lossless, LZW, RLE or Zip as alternate compression methods (some only allow b/w images). For details see here.

I was able to achieve great compression ratios for scanned/photographed documents (depending on the settings). Depending on the document source, you might want to reduce the color depth (-depth argument).

  • 3
    For a scanned document where the text is what you are interested in rather then the images and preserving depth isn't an issue, jpeg compression is not a good idea because the artifacts tend to be extremely noticeable. If you use pdfimages input.pdf pages to extract pbm files, then you can do something like: for page in *.pbm; do convert $page -compress Group4 -type bilevel TIFF:- | convert - output.pdf. Any OCR will be lost so I usually then do pdfsandwich output.pdf, which seems to reduce file size even further. – Brian Z May 4 '15 at 11:57
  • 1
    @BrianZ sure jpeg compression isn't always the best choice, but for me it was the best approach for mixed type documents. I added some informations about other compression methods to the answer. – someonr May 6 '15 at 23:43
  • 2
    This method ultimately uses gs behind the scenes. – alfC Jun 12 '15 at 4:55
  • 1
    I had to use double dash for the options to run the command --density --quality --compress vs -density -quality -compress. – Rotareti Nov 10 '16 at 18:22
  • 1
    If image qaulity is not the highest concern (and you just want to get that dang email attachment small enough to be sent) one might add -resize 50% too, change percentage depending on how much DPI was used while scanning – chrki Jan 11 '17 at 0:45

I needed to downsize a PDF that contained full color scans of a document. Each of my pages was a full color image as far as the file was concerned. They were images of pages containing text and images, but they were created by scanning to an image.

I used a combination of the below ghostscript command and one from another thread.

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dDownsampleColorImages=true \
-dColorImageResolution=150 -dNOPAUSE  -dBATCH -sOutputFile=output.pdf input.pdf

This reduced the image resolution to 150dpi, cutting my file size in half. Looking at the document, there was almost no noticeable loss of image quality. The text is still perfectly readable on my 2012 Nexus7.

  • 2
    +1 for down sampling images but keeping text as vectors. Made a huge difference in side without making my text pixelated. – Jason O'Neil Dec 8 '14 at 8:34

Here is a script for rewriting scanned pdfs:

#!/bin/sh

gs  -q -dNOPAUSE -dBATCH -dSAFER \
    -sDEVICE=pdfwrite \
    -dCompatibilityLevel=1.3 \
    -dPDFSETTINGS=/screen \
    -dEmbedAllFonts=true \
    -dSubsetFonts=true \
    -dColorImageDownsampleType=/Bicubic \
    -dColorImageResolution=72 \
    -dGrayImageDownsampleType=/Bicubic \
    -dGrayImageResolution=72 \
    -dMonoImageDownsampleType=/Bicubic \
    -dMonoImageResolution=72 \
    -sOutputFile=out.pdf \
     $1

You could customise it a bit to make it more reusable but if you only have one pdf, you could just replace $1 with your pdf filename and bung it in a terminal.

  • 1
    Works a treat, thanks Oli. You've answered pretty much everything I've asked on here so far :-D – Rob Cowell Sep 1 '10 at 8:15
  • This is a good answer but in my case at least it takes a lot of time to convert a somewhat large (>10Mb) PDF file (more than a minute). – Gabriel Jun 12 '13 at 19:20
  • I'm not sure what happens, but a 30 MB PDF results a 68 MB file. Instead of reducing, it enlarges. Same output if using directly ps2pdf as stated in next answer. – Ed Villegas Jun 23 '13 at 18:08
  • @EdVillegas The only thing I can think of (to explain that sort of increase) is that the images are of a lower resolution than the ones being generated (72dpi). Or somehow embedding the fonts is sucking in all the fonts. – Oli Jun 25 '13 at 7:31

I usually use ps2pdf to do this (easier syntax), something like this:

ps2pdf -dPDFSETTINGS=/ebook BiggerPdf SmallerPDF

I use the following python script to reduce the size of all the pdf files in a dir in a production server (8.04). So it should work.

#!/usr/bin/python

import os

for fich in os.listdir('.'):
        if fich[-3:]=="pdf":
                os.system("ps2pdf -dPDFSETTINGS=/ebook %s reduc/%s" % (fich,fich))
  • Thanks for the alternative solution. I tried Oli's first and it gave me the result I needed, but I will keep this one for future reference too. – Rob Cowell Sep 1 '10 at 8:17
  • -dPDFSETTINGS=/printer option do 50% resize. ebook do 90% resize. – neouyghur Apr 13 at 6:08
  1. I use LibreOffice Draw to open the pdf.
  2. I then "export as pdf"
  3. And set "jpeg compression quality" to 50% and "image resolution" to 150 dpi

This will have a good result.

  • Worst possible solution to the problem! It totally messed my file up! – ThatIs Jan 15 at 16:31

Best for me was

convert -compress Zip -density 150x150 input.pdf output.pdf

Other ways:

#### gs
gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/ebook -dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf $INPUTFILE

### pdf2ps && ps2pdf
pdf2ps input.pdf output.ps && ps2pdf output.ps output.pdf

### Webservice
http://compress.smallpdf.com/de

regards

  • 1
    good summary thanks – Gery Jan 28 at 20:10
  • Great solution. Easy to remember and it brought my pdf from 32 to 3.5 MB without sensible loss in readability. – Immanuel Weihnachten Mar 17 at 9:57

I just encountered this problem myself. If using simple scan, select text mode for low resolution scans and you won't need to worry about the command line stuff. Just saying.

  • 1
    This is the single answer in this thread that solved my problem. I downplayed Simplescan, but it really was the answer for me, instead of fighting against Xsane in what seemed to be an endless agony. – versvs Aug 31 '15 at 16:03

Since this link was first for me when I searched in Google, I thought I'd add one more possibility. None of the above solutions was working for me on a pdf exported from Inkscape (15 mb), but I was at last able to shrink it down to 1 mb by opening it in GIMP, and exporting as pdf again.

Another option that came close (but text was a little fuzzy) was ImageMagick's convert utility:

convert -compress Zip input.pdf output.pdf
  • I guess this is what you meant by "a little fuzzy", but just to clarify, convert -compress Zip appeared to rasterise all vectors. – Sparhawk Feb 22 '15 at 3:39

In the end I wrote my own bash script to solve this, it uses mogrify, convert and gs to extract pdf pages as png, resize them, convert them to 1-bit bmp and then rebuild them as pdf. File size reduction can be over 90%. Available at http://www.timedicer.co.uk/programs/help/pdf-compress.sh.php.

Control the compression quality:

#!/bin/sh
INPUT=$1; shift
OUTPUT=$1; shift
GS_BIN=/usr/bin/gs
QFACTOR="0.40"

# Image Compression Quality
#
# Quality HSamples VSamples QFactor
# Minimum [2 1 1 2] [2 1 1 2] 2.40
# Low     [2 1 1 2] [2 1 1 2] 1.30
# Medium  [2 1 1 2] [2 1 1 2] 0.76
# High    [1 1 1 1] [1 1 1 1] 0.40
# Maximum [1 1 1 1] [1 1 1 1] 0.15 

${GS_BIN} -dBATCH -dSAFER -DNOPAUSE -q -sDEVICE=pdfwrite -sOutputFile=${OUTPUT} -c "<< /ColorImageDict << /QFactor ${QFACTOR} /Blend 1 /HSample [1 1 1 1] /VSample [1 1 1 1] >> >> setdistillerparams" -f ${INPUT}
  • ...so both INPUT and OUTPUT are the same argument? You might want to add usage guidelines. – mikewhatever Apr 2 '16 at 11:46
  • 1
    Note the shift. First parameter is input file, second is the output file and rest of the parameters will be passed to gs as is. – Mikko Rantalainen May 13 '16 at 12:53
  • I think you want /HSamples and /VSamples, not /HSample and /VSample. See e.g. ps2pdf docs or the PostScript language reference manual. Also perhaps worth noting that the allowed QFactor range is 0 to 1,000,000 and lower values produce higher quality. – Pont Aug 2 '17 at 8:31

If converting to djvu would also be ok and if no colors are involved, you could try the following:

Convert the pdf to jpg files using pdfimages -j

If you get pbm files instead, you should do the intermediate step:

for FILENAME in $(ls *.pbm); do convert $FILENAME ${FILENAME%.*}.jpg ;done

The convert command is from the imagemagick package.

Then use scantailor to make tif's out of it.

In a last step you go to scantailors out direcory (where the tif's are located) and apply djvubind to that directory.

This should reduce the filesize drastically without big quality loss of the text. If you want finer control over the ocr-backend, you may try djvubind --no-ocr and use ocrodjvu to add the ocr layer afterwards.

If you have color's in your document things get a bit more complicated. Instead of djvubind you could use didjvu and in scantailor you have to change to mixed mode and select sometimes color-images manually.

load image or even pdf file into inkscape.

From inkscape: Save in vector format (as the native .svg).

Import vector files into scribus, edit layout and export/save as .pdf from there

I was facing the same problem, and was glad to find this thread. Specifically I had a pdf generated from scanned images, and needed to reduce its byte size by a factor of 6.

Unfortunately, none of the solutions above worked :(. Then I realized that somewhere in the scanner->jpeg->pdf process the size of the page had gotten bloated by a factor of aprx 4. The documents I scanned were all Letter sized, but the pdf had size of

identify -verbose doc_orig.pdf | grep "Print size"
 Print size: 35.4167x48.7222

I got the desired results finally with a "convert" command that did both resizing as well as compression steps in one:

convert -density 135x135 -quality 70 -compress jpeg -resize 22.588% doc_orig.pdf doc_lowres.pdf

Note that doc_orig had density of 72x72 dpi.

Super simple PDF compress tool: GitHub page.

Installation on Ubuntu:

sudo add-apt-repository ppa:jfswitz/released

sudo apt-get update

sudo apt-get install pdf-compressor

It uses ghostscript.

I recently discovered and strongly recommend pdfsizeopt.

It is much more efficient than any of the previous CLI and GUI software that I have tried (including convert, gs, pdftk, etc.) — although possibly slower with pngout activated — and does not have some of their issues (no heavily pixelated/degraded images, no loss of bookmarks, etc.).

I recommend the generic Unix install:

  1. Install all required dependencies:

  2. Download and install the executable:

    curl -L -o https://raw.githubusercontent.com/pts/pdfsizeopt/master/pdfsizeopt.single
    cp pdfsizeopt.single /usr/local/bin/pdfsizeopt
    

Usage:

pdfsizeopt original.pdf [compressed.pdf]

You can try this :

$ time pdftk myFile.pdf output myFile__SMALLER.pdf compress
GC Warning: Repeated allocation of very large block (appr. size 16764928):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 8384512):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 11837440):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 8384512):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 33525760):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 7254016):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 34041856):
    May lead to memory leak and poor performance.
GC Warning: Repeated allocation of very large block (appr. size 33525760):
    May lead to memory leak and poor performance.

real    0m23.677s
user    0m23.142s
sys     0m0.540s
$ du myFile*.pdf
108M    myFile.pdf
74M     myFile__SMALLER.pdf

It is faster than gs but compresses upto 30% in this case for a 107.5MiB input file.

I used below commands but it didnt compress my pdf file substantially. Some times some of the portion was blackened after compression.

  1. gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/ebook -dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf $INPUTFILE

  2. "ps2pdf -dPDFSETTINGS=/ebook %s %s" % (input_file_path, out_file_path)

After too much wandering over the web I just couldn't find the right compression library. I came across pdfcompressor.com. This is just awesome website. It compresses the pdf by 95% ( 15Mb of files). So I used selenium and Tor to automate the compression. Checkout my Github Repository. [GITHUB] (https://github.com/gugli28/PdfCompressor)

protected by Community May 9 at 1:52

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.