Choosing Graphics formats

Article contributed by John McGhie

 “What’s the best graphics format to use in my document?”

Use JPEG for photos, PNG for screen-shots, EMF for drawn pictures, WMF for simple drawn pictures, and EPS and TIFF for professional publishing.  Use RGB colour format (24-bit colour) always, except for professional publishing.

There: now you have your answer, you can go about your way without cluttering your mind with useless information.  Unless you want to know “why” I say this!

The following links have in-depth information you may be interested in.

Why you might need to know

Ever since there have been “people” there have been “pictures”.  Pictures came well before “writing” and thousands of years before “computers”.  So this is actually a very old question!

The answers from yesteryear are no longer so relevant today.  Mud-maps drawn with a stick in the dirt have the advantages that the software to create them and display them is very cheap, learning to use it takes no time, and any human being can decode the resulting image.  However, portability is an issue, and the first shower of rain can affect the permanency of the image.  Not a good format for displaying colour, either.  Carving on stone tablets scores highly for permanency, but the tools to create pictures become more expensive, the technology is more difficult to learn, portability remains an issue, the ability to revise the work is severely limited, and the colour gamut is severely restricted.

Notice how as soon as we raise this subject, the discussion wanders unbidden down a track that could be described as “what do you want to do with the result?”  That’s a useful learning, and it’s the key to the whole puzzle.  How do you want to USE your picture? Who do you want to use your picture?  WHERE?

Let’s define two kinds of graphic:

And four purposes:

Now we can discuss the constraints that each kind and purpose imposes.

Raster Graphics

I would much rather you headed off to Wikipedia to read a proper definition of a raster graphic.  Since I suspect you won’t, I’ll tell you: it’s an image made up of a pattern of dots, like a TV picture (go up and stare at your TV screen from close enough to see the individual dots if you don’t believe me…).  Here’s a proper description.

Raster graphics are generally best for handling subtle variations in colour (e.g. photographs). Their extreme disadvantage is that they are of fixed size: if you try to alter the size of a raster picture it gets the jaggies.  If you try to increase the size to make it print at the same size as it appears on screen, it often looks really blocky.

Advanced graphics software has special tools that enables you to produce a fairly good result when scaling raster graphics.  Word is not such software.  Moreover, you don’t get anything for nothing in this world, and you won’t get good print quality from a raster graphic unless you produce it at the size you want to print it.

Vector Graphics

If you wouldn’t go to Wikipedia to read about raster graphics, is it safe for me to assume that you won’t follow this link to read about vector graphics either?  Sigh… 

A vector image is composed of mathematical formulae.  Each formula describes a shape.  Really there are only three shapes that we need to be concerned with: lines, polygons, and circles.  Every other element in a picture can be described by a combination of and variation upon those three basic shapes.

Vector graphics are generally the best for printing, particularly for printing “simple” pictures such as those we draw.  You can scale a vector graphic to any size you want, just by dragging a corner.  It always prints or displays at the maximum resolution of the display or printer, it can never get the jaggies.

The disadvantage of vectors is that each format is subtly different: there is as yet no wide agreement among software vendors as to how to express these files.  Consequently there is always a risk you will find a device or piece of software that can’t handle a given format.

There is a sub-group of vector graphics known as “metafiles”.  Metafiles don’t actually contain the formulas, just the names of the shapes and their measurements.  The benefit is that each operating system has its own native metafile format.  When producing pictures for a specific platform, use its native metafile and you get perfect reproduction.  The native format for Windows is EMF (Enhanced Metafile).  The native format for Mac is PDF.

In professional printing, the advantages of using metafiles far outweigh the disadvantages.  For everything else, it’s a judgement call.  I would recommend that you always try to use EMF or PDF for drawn illustrations until you find a reason why you can’t.


Documents you wish to email must be small.  Anything over 250 kb is going to annoy most recipients.  Anything over one megabyte is going to be rejected by the mail server at most people’s workplace.  Anything over ten megabytes is going to overflow people’s mailboxes, shut down their email accounts and cause them to add your email address to the “Junk Mail Senders” list, from which your emails will never again be seen.

The other issue you have with email is that you may not know what kind of computer your recipient is using, or what software is installed on it.  You have an 80 per cent chance that it’s a Windows computer: but there’s a ten per cent chance that it’s a Mac, and ten per cent chance that it’s “something else” -- like a mobile phone.

Businesses these days are increasingly paranoid about their mail servers.  The mail server is mission-critical to every business or government organisation these days: they can’t afford to let a virus loose in it.  So mail administrators have changed from “rejecting known baddies” to “trashing everything we can’t prove is harmless.”

To guarantee that your email is received, don’t send rare formats (no matter how harmless you know them to be) and keep them small.  Use JPEG, WMF, EMF, or PICT.  Don’t try PNG or GIF because they’re known to be able to contain viruses.


On the web, we would like pictures to be small but this is no longer an absolute rule.  The Internet has improved a lot, and even simple web editors enable you to provide alternate “small” and “high-res” versions of each image.

For general Internet use, stick to JPEG and PNG to give you the broadest compatibility with all the devices that might be accessing your page.  Use JPEG for photos, and PNG for screen-shots and drawings.  Both of these formats compress the picture information to speed download time.

JPEG removes detail to preserve colour.  So always use JPEG for “continuous tone” pictures such as photos: face tones will look natural but some of the fine detail will be lost (which can be an advantage when you get as old as me!).  Most JPEG editors enable you to specify how much detail you are prepared to lose.  Adjust the quality setting, save, and examine the result to see if it suits your purpose.  As a suggestion, start with a setting of around “50 per cent” or “medium” quality for use on the web.  If you turn the quality down so the loss is just noticeable, JPEG is about half the size of PNG. 

PNG is a relatively new format.  It is a non-lossy compression, which means you do not lose any image detail, and it’s royalty free because it’s not covered by patents.  If you think your user would like to print the picture, use PNG so when they do print, they haven’t lost any detail.

PNG replaced GIF as the best format for web graphics.  The vast majority of browsers these days can display and print PNG, and it is smaller than GIF.  However, some devices such as mobile phones will support GIF but may not support PNG.  If your images may be used on mobile phones, and you are prepared to compromise the appearance of your site to ensure that the less than one per cent of users who cannot view PNGs can view your site, you may choose GIF.  GIF removes colour information to preserve detail. GIF allows only 256 colours per image: which means you will get strange banding effects on continuous-tone images (so don’t use it for photos).  For screen-shots or line-drawings, it’s fine.  If you need more colours, use PNG.

JPEG, PNG and GIF all allow “progressive” downloading: they send every tenth line, then every fifth, etc so the picture appears quickly in low resolution and then fills itself in.  GIF also supports animation.  Word does not support progressive images, and it does not support animated images

PNG and GIF also support "transparency": enabling you to specify that one of the colours in the image be rendered as transparent.  If you use transparency you need to check the result in your target browser: some may have issues with transparency in PNG.  All the graphics on this page are in .PNG format.  If you look very closely at the corners of the MVP logo at the top right, you might see that they're white and square instead of blue and rounded.  If you see this, the browser you are viewing this page with has an issue with transparency in PNG.

Compatibility and appearance are the keywords on the web.  For this application, your pictures must look good (that’s the whole purpose of putting them there) but they must appear on any recipient’s device.  You still have no idea what that device is, but you can make some assumptions.  For example: you can assume “a computer” and “at least 800 x 600 pixels display capability”.  If you want your website to look good on a mobile phone, you need to do a whole lot of things that are beyond the scope of this article, but for starters assume a 128 x 128 pixel screen size and no more than 256 colours.

On the web, avoid vector-format graphics of any kind (including Flash).  Despite the best efforts of the community there is still no vector graphics format you can guarantee to be available on every device that accesses the web.  When there is, we’ll let you know!  Despite the impressive ability of many web design houses to extract millions of dollars from major corporations to prepare “Oh Wow!” Flash graphics and animations, most web users find them unbelievably annoying, and a large percentage of users have Flash turned off (because it can be a security risk) so they will see none at all.

If you are in a known environment (for example: at work where the company specifies which software and browsers everyone will use) you can relax the rules slightly.  On the corporate network, “bandwidth” is like air: it’s good for you, breathe deeply!  You can allow your file sizes to run up as high as you need to accomplish your requirement.  Just be aware that if your finished graphics are often larger than a megabyte you are likely to generate a “please explain” from your system administrator!  And find out before you start whether you have any users working from home: if you do, and you create huge graphics, you will hear from them, and you may wish you hadn’t!

In an all-Microsoft environment, you can use both WMF and EMF on web pages, and the result will look much better and print much better than GIF.  If you are using Visio to prepare your graphics, you can also use WMZ and EMZ.  These are compressed versions of WMF and EMF respectively that are also “scalable” in the browser.  You no longer need to know what the user’s screen size is: the picture will scale to fit.  You will need to read the Visio help carefully and do a few practice runs to work out how to do this, but the result makes it worth persisting.  However, be warned: the resulting pictures/web pages can be displayed correctly ONLY on a Windows XP PC in Internet Explorer.  So if you are not in a known, fixed environment, don’t try it.

Locally-printed documents

When preparing documents to be sent to a printer attached to your computer or on the office network, you can let yourself off the file-size leash.  While it is still possible to create pictures that are too large to print, most companies these days specify their printers well enough that this will rarely be an issue.

Word will cope quite happily with huge graphics file sizes: the latest versions will allow up to two gigabytes of graphics in a document.  In a Word document, the graphics and the text are stored in different sections of the file.  Provided that the text portion of the file does not exceed 32 megabytes (about 5,500 pages) or the entire file does not exceed two gigabytes, Word will open and save the file.

However, once the document goes above a hundred megabytes, a normal computer gets very, very slow.  Word is not the problem, nor is the CPU.  The problem is that the hard disk can’t read and write fast enough, and the machine may be running out of memory (RAM).  There are two things you can do:

The first is to use graphics that are external to the file.  Look up the Word Help topic “Insert a picture” to see how to insert a picture as a linked graphic.  This technique means Word doesn’t store the graphic in the document, which means you can edit and save dramatically faster.  When using linked graphics, create a folder for the document and make sure that the document and its pictures are all in the same folder.  This enables you to zip the folder up and move it to someone else’s machine without breaking all the links.  Better: put the folder on the network and tell other users where it is, instead of moving the document.

The second response to slow documents is to demand a serious workstation.  If you spend your days creating large documents full of high-quality pictures, the company should think about investing in a proper graphics workstation for you, so you can work efficiently.  For around $15,000 you can get a dual-CPU machine with four gigabytes of memory and a fast SCSI disk array.  Such a machine will run perhaps a hundred times faster than a standard corporate PC on long and complex documents (but it’s a waste of money for ordinary office documents).


You can use any file format that is common at your office.  However, there are two “gotcha’s” you must be aware of:


In an office setting these days, nearly every printer you come across will be able to handle Adobe’s PostScript printer page definition language.  But if you are working in the one office in a thousand that cannot support PostScript printing, you must not include EPS or AI graphics in your documents, or they won’t print!

If you do not know whether or not your printers support PostScript, ask your system administrator.  Ask very delicately, because if the answer is “no”, someone who should have known better made a serious mistake in ordering, and you may be talking to that person!

Sometimes, your system administrator needs to install a special printer driver on your computer to make the printer’s PostScript features available.

Special-purpose software

If you prepare a lot of graphics, you will have requested (or been issued with) a proper graphics package.  Be aware that the ordinary Microsoft Office users surrounding you cannot display many of the formats it can produce.

If you copy an Adobe Illustrator picture and paste it into a Microsoft Word document, that document will display perfectly on YOUR computer where Word can start Adobe Illustrator in the background.  Everyone else will see and print a blank square.  Same with Visio or CorelDRAW.  Just because you have it on your machine does not mean everyone else in the company has.

As a graphics professional with a keen desire to remain employed, it’s your responsibility to ensure you do not leave the company knowledgebase full of documents that no-one but you can display or print.

Save your pictures in a compatible format: WMF for preference.  If you need more colours or rotated text, you can use EMF, but be aware that Macintosh computers cannot display or print it.  You can save to PDF or PICT in a mixed environment, but be aware that PICT requires other users to have QuickTime installed.  If they haven’t, use PDF or PNG.

Professionally-printed documents

Documents for professional printing are typically prepared (or at least finished off…) by a graphics professional whose job it is to know all this stuff.  Normally a commercial printing shop will expect the document delivered in PDF format at “Prepress” quality.

For this you will need a copy of Adobe Acrobat software.  Microsoft has announced that they will make a download available for Office 2007 that provides the ability to save documents in PDF format.  This utility may not include the professional-level features required to adjust the quality.  The quality setting in PDF has no effect on text: the text is produced from the installed fonts at the maximum resolution.  The quality setting adjusts the resolution of the included graphics.  If you do not have a copy of Adobe Acrobat, check around in your company: it’s expensive, so typically a company has only a few licences. 

For graphics professionals

All we need to tell you is “Don’t go above 24-bit RGB and don’t try CMYK or Pantone in graphics destined for use in Microsoft Office documents.

Microsoft Office is generally limited to RGB, eight-bits-per-pixel (thousands of colours, on a Mac).  Office 2007 is rumoured to be getting CMYK support, but we will wait until it ships before claiming that it has.

If you are working on a Macintosh, make sure you use only the fonts Microsoft supplied with Office, or be prepared to deal with the fact that parts of your picture will move around when displayed or printed on a PC.  Note: it’s not the “name” of the font that counts: Microsoft contracted its font vendors to prepare a special batch of fonts for Microsoft Office on the Mac.  These fonts are carefully encoded to match the fonts with the same names that are shipped in Microsoft Office on the PC.  If you use other fonts, the metrics will be off on the PC, and so will the encoding: all PC fonts are Unicode, only some Mac fonts are, so you will get some strange characters.

For the rest of us

Assuming you are not a graphics professional (if you were, you probably wouldn’t have read this far…) we need to open the door to the very complex world of professional publishing for you.

I get annoyed by articles that try to make things “simple” or “easy” by leaving out most of the information I came to get.  So let’s plunge right in: there’s a lot of technical detail ahead, and some of it’s not easy.  Just remember: the person who gave you the job knows you’re clever enough to complete it.  I know you are.  If you don’t believe us, you’re wrong!

You first need to define the goal you are trying to achieve.  “Professional” means “conforming to specifications”.  One of those specifications is “cost”.  If your job costs too much when you get to the print room, it’s going to get cancelled, you are going to get yelled at, and we will all miss our deadline.

The three things that can make or break your print job are: colour, paper stock, and resolution.


Colour is the biggest and deepest hole your project can fall into…  It is a very complex topic, about which many books have already been written.  I'm going to skip lightly over the high-points here, mainly to give you the substantiation you need to convince your boss that you may need to retain a professional once you decide to print a large job in colour.

Each different ink you need requires a trip through the printing press for each side of each sheet that you print, plus set-up and wash-up time between each ink.

Let’s assume you have a book of 500 pages you want to print.  To run that job in monochrome (black and white) it will cost you around three cents per impression (two impressions per sheet x 250 sheets + paper and binding = $15.00 a copy) in lots of 50.

To print “Spot colour” (e.g. black text and blue headings) will cost twice that: $30.00 a copy.

To print in “process colour” will mean four trips through the printer (cyan, magenta, yellow, black), and four lots of set-up and wash-down = $60 per copy.

But unless you hose down the idiot in Corporate Branding that wants to use three Pantone colours in the company logo that occurs in the running header or footer, that’s going to take another three trips through the printer, adding another $45 per copy.  So your tender proposal that was going to cost $150.00 to print the required ten copies suddenly ended up costing $1,050.00.  Just to print it.

So as a rule of thumb: to go from black and white to colour quadruples your printing cost.  But each Pantone colour is a separate ink, which means another trip through the printer.  You cannot reproduce accurate Pantone colours using CMYK process inks.  However: you can come close.  Quite close.  Tell the guy who wants to run Pantone colours to produce a business case showing how many more sales the document will generate for the extra cost.  Then show him your test prints done in process-colour, and ask him to re-run his business case based on the colour variation.  If you get a skilled graphics professional on the job, they can often re-work that logo so it does not require Pantone colours.

Commercial printing in colour can require specialist skills, and functions that are beyond the capability of Microsoft Office products. No Microsoft Office products yet provide the ability to separate a print job into its respective colour separations.  Depending on the colour process you choose, you may need to have this step performed by the commercial printing company.  Expect them to charge for the service: producing a high-quality result is a skilled job.

Paper Stock

Once you have decided on your colour process, your printer (person) can guide you on selection of paper stock.  Some colour options require specific paper stock.  High-resolution colour photos requires heavyweight glossy stock.  Recycled paper may require special inks.

Generally it’s cheaper to print on large stock and guillotine than it is to print on the correct-sized stock.  If your publication can be printed four-up or eight-up the savings in handling time and job run-time may be very considerable.  Ummm…  make sure the printing company reflects these savings in their quote!

Once you have chosen your paper stock, the printer will order stock in.  Before committing, run some test prints and get Marketing, Sales, Corporate Branding and Product Engineering to sign off on the test prints.  Because it may not be possible to change your colour process once you have ordered the paper, without paying for the paper you’re not going to use.


In professional printing, resolution is like sex or money: it’s only a problem if you don’t have enough!

Actually, that’s not quite true.  It is technically feasible to produce images of such complexity, size and resolution that they will crash the software when you attempt to make up the job.  This is very unlikely to happen these days, and if it does, you may wish to look for a more capable printing firm.

Most commercial printers will do a grand job with originals that reach at least 600 dpi for black and white and at least 150 dpi for four-colour.  Your printing firm will guide you as to how they want your images prepared: and they may advise very specific requirements, depending on the process they intend to use.  Generally, the bigger the job and the higher the print-run the more money you will save by doing things exactly the way the printer wants.

In the absence of any other suggestions, produce your vector images in EPS and your raster images in CMYK TIFF.  It is usually OK to compress the TIFF images, but check.

Within limits, EPS has unlimited resolution, but only for the lines and shapes.  Many people do not understand this: EPS is a “capsule”.  It certainly contains shapes, expressed in PostScript.  But it also can contain raster images.  This depends a lot on whether the software that wrote the EPS was capable of translating the shapes in your drawing into vector formulae.  For the simple ones, it can.  For things such as blended fills, less expensive software may fall back to producing a bitmap (a raster image) of the shape.  Visio has a bad habit of doing this…

For raster graphics, make the resolution at least 150 dpi for colour, at least 300 dpi for half-tone, and at least 600 dpi for line art.  However, don’t get it too high.  Not only is there no point (the human eye can’t see the difference…) but it can slow down the print run (double or triple the time) and may exceed the capabilities of the output software (which will then crash!).  Modern digital cameras can produce very high resolutions: don’t get greedy!  I would not go above 600 dpi for colour, 900 dpi for half-tone, and 2400 dpi for line art without testing very carefully.  There is no point in going above 1,200 dpi unless you are printing on glossy stock: you can’t see the difference.  Many medium-grade commercial machines can’t produce images with an actual resolution above 600 dpi.

I would prepare at least 12 pages of test material to test with: any machine can print one page, but what happens when it has to keep making up pages one after the other for a long run?

Be aware that if you produce CMYK TIFF images, it is possible to produce colours that a process-colour printing press cannot reproduce.  Print some test pages and try them out, at least a month before you want to run the print job.  In professional printing “If you don’t KNOW, the answer’s NO!” So the time to have a very specific discussion -- not with the clerk on the sales desk, but with the printing foreman who is going to run your job -- is at LEAST a month out from the press date.  Right?  After all, if it turns out that they simply cannot run your job on their equipment, you need time to find another company and re-book the job.


Don’t fall for the old “You must produce this on a Macintosh and supply all the fonts you used” nonsense.  If they want your money they better find a way to print what you brought.

Your author has very successfully produced long-run books of 5,000 pages fired directly out of Word on a PC into the printer’s equipment.  If your printer can’t (or doesn’t want to…) do this, find another company.  Most printers have enough profit-margin in a reasonable corporate job to go out and buy a brand new PC just to run your job, if they do not have one already.

I might add that the reason I ended up going “direct to plate” from Microsoft Word was that the PC I was using at a customer premises lacked the memory to get Adobe Acrobat to complete the PDFs.  Please be aware that Acrobat requires prodigious amounts of memory (over a gigbyte) to successfully convert a large and complex Word document to hyperlinked PDFs.  If you haven’t got enough resources in the machine you’re using, it will just hang and you’ll get nothing.  Don’t find this out like I did -- at eight p.m. with the print run scheduled to begin at ten.

If this happens to you, you can work around it two ways: either print from Word directly to the Adobe Acrobat PDF Printer (without using PDF maker), or install the appropriate printer driver for the printer’s output device on your PC (or Mac!) and print to file using that driver.  Printing directly to the PDF Printer will produce a PDF file, but there will be no hyperlinks or TOC in the file.  You would then produce two PDFs, one at pre-press resolution for the commercial printer, one at Internet resolution with hyperlinks and TOC.  Make the first by printing direct to the PDF writer, the other by using PDFMaker.

Printing directly from Word is the other option.  You must either take the document file to the printer’s premises, or install the correct printer driver for their equipment.  If in doubt, almost anything will work with the Apple LaserWriter 600 driver that comes standard with Windows.  Don’t forget to specify the correct paper size!

Use File>Print in Word and choose the correct printer driver.  When you change printer, the Cancel button name changes to “Close”.  Click “Close” (so the job doesn’t print yet).  This changes the measurements that Word reads from the printer driver for things such as the paper size, the printer resolution and the font measurements.  Click CLOSE, then and go through and carefully check your pagination from front to back for the entire book.  When you switch printer drivers, the document WILL reflow.  You can’t stop it.  If you have formatted correctly the way we recommend, the document will reflow perfectly.  If your document is full of blank lines and hard page breaks, you will get to fix it all!

Now use File>Print.  If you are doing this at your office, check the “Print to File” box and click OK.  Word will prompt you for a file name and folder.  Create a name for the file and save it someplace where there is plenty of room.  The system will produce a raw PostScript file: these are huge: a one megabyte Word document will turn into at least 24 megabytes as a PostScript file.  You will need to Zip the resulting file and carry it to the printing company on a data key or hard disk.

From the archive

This article was re-created to replace the article that was here before your hapless webmaster hit the wrong button and deleted it.  We managed to retrieve some of the original (Google is your friend...).  The bits that are still relevant to modern practice are included below.

Online versus paper formats

For information to appear in online formats, it needs to appear on a screen or display device.  Ignoring for a moment the complexities of how it got there, there is an absolute limit to how well the display device can display an image.  Every real world display device has limits, but the ones that you and I can afford are surprisingly poor.

The majority of computer screens are absolutely limited to 96 dots-per-inch, and to a contrast ratio (the brightness difference between the blackest black and the whitest white) of about 100 to one.  The fancy screens used in graphics pre-press will do a little better but not much.  LCD screens are notorious for having a poor contrast ratio: 70 to one or thereabouts.

For information to get on paper, it must first be "printed".  Sounds obvious, until you begin to understand the complexities involved in printing computer information on paper.  So let's ignore them for a minute and consider this: if the printer was perfect, then the limit of quality on paper is the quality of the human eye.  So just how good is the human eye?  Debate rages... I encourage you to look this up on the web.  However, let's assume for a moment that we are talking about a picture that consists entirely of black lines on a white background: the theoretical limiting resolution of the human eye is about 3,400 dots per inch.  You would need very, very young eyes to see that well.  Most folks won't see the difference much above 350 dpi.  Our resolving power in colour is nowhere near as high.

Now here's where the complexities come in.  I am going to skip lightly over the worst of it.  If you really want to know, it's out there on the web: go look.

On screen, every pixel stays the same size.  One pixel is one pixel: each pixel is made up of three spots, red, green, and blue.  Each spot can vary in brightness.

On screen, every pixel can have a brightness and a colour.  On paper, it can't: it is either black or white, red or white, green or white... depending on the number of inks you have in your printer.  You either have "ink" or "no ink" on that spot on the paper.  To print "shades" and "colours", a printer combines several pixels to achieve the effect it wants.  Ignoring completely the mathematics involved, it takes about ten printer dots to achieve one pixel of printed image.  Modern printers can vary their dot sizes to produce lighter or darker shades.

On screen, colours are produced by adding colours together.  On paper, colours are produced by absorbing every other colour from white light, reflecting only the "colour" of the ink.  On screen, you have only red, green, and blue: you produce a colour by altering the brightness of each of these three colours in relation to the other two.  On paper, you vary the darkness of each of four colours to subtract light until what you have left is the colour and brightness you want.

On screen, each pixel can be any colour, and you can combine as many colours as you wish in a single pixel.  On paper, each spot can be only one colour: the colour of the ink dot that hit it.  Modern printers hold up to six inks, but they can't all land on the same spot.  They can get very close to each other, but not on top of each other.  The manufacturer's challenge with a printer is to get the dots of different colours as close as possible to each other.

Printing presses, on the other hand, can put multiple colours one on top of the other.  You see, a printing press runs the paper through once for each colour, and can do this as many times as you wish to pay for.  However, disaster strikes if the colours actually do overlap where you didn't want them to.  You then get a very obvious line of a different colour at the border between each object.  Your challenge, as a graphics professional, is to know how closely you can place objects of a different colour without the inks running into one another.  The actual measurement varies depending on the paper you are using, the individual inks you are using, and the humidity of the paper on the day.  The reason graphics professionals spend years at university is to ensure that their first guess is close to being right.  The rest of us learn to leave a one-point "trap" around an object and hope. A trap is a gutter of blank paper left between two areas of different colour so that when the ink soaks into the paper if doesn't touch the neighbouring area. Your printer will gleefully inform you if you get it wrong: he or she can then charge you a serious fortune to prepare new colour separations with adjusted trapping widths, or an even more serious fortune to run the job through the press multiple times, waiting for the ink to dry between each impression.

In an ideal world, you would capture as many colours and pixels as you could, and send them all to your display device or printer.  You would then enable the display device to choose the information it can use and ignore the rest.

In the real world, if you do this your readers will beat a path to your door all right, but it will not be with offers of a higher salary.  More like offers of unemployment.  The hardest message to get across in professional publishing is that "perfection is completely unacceptable."  Nobody has the time, bandwidth, disk space or budget to get it perfect.  As a documentation professional, your job is to ensure that what you produce is at the minimum quality needed to fulfil the publication purpose.  Anything more will rapidly consume resources, time, and your employment prospects.

The highly-thought-of professional produces just enough to do the job: no more.


Let's do resolution first, basically because it is easy to understand.  Here, the rule is "Capture all you can, publish as little as possible."

If you are on site with a film camera, burn film.  With a digital camera, set the resolution and number of colours as high as you can get and bang away until you fill the memory.  Ten rolls of film might cost you fifty bucks when processed.  How much is it going to cost to organize all the people, travel to the venue, wait for suitable weather and set up everything again?  Now, we can understand that you didn't get the shot you needed because you tried to save film, but your boss's comprehension may not be that good.

You can always remove information from the file when you get back to your desk: but you can never add any that you did not capture in the first place.

For photographs and continuous-tone material, choose TIFF for professional publishing, PNG for office printing, and JPEG for the web.

If you are publishing technical drawings and line art, make sure you produce and publish in a vector format.  Inexperienced authors who try to print line drawings from a raster are unlikely to get the opportunity to gain much more experience: it looks really disgusting.

For Publishing you must use EPS — publishing houses can't (won't!) cope with anything else.  For office-printed output use EMF or WMF.  For the web, there are no good vector formats that are widely available.  For more information, see The Formats and  Electronic Output further on in this article.


Reduce this as much as you can.  While the human eye can handle a contrast ratio of millions to one, no electronic or paper medium can.  Learn to look critically at your pictures: can you see detail in the shadows?  Can you see detail in the bright areas of the picture?  If not, your contrast is too high.

Does your picture look washed-out or dull?  If it does, your contrast is too low.

Number of Inks

If you find a way to change this on the web, Adobe would like to hear from you.  Anything going out to a display device is limited to red, green and blue.

If you are going out to paper, learn to minimize the number of inks you require to accurately print your picture.  For a laser printer, you cannot do much about it.  Most lasers are black-and-white or four-colour and that's what you have to work with.

If you include photographs in a document you intend to print in black-and-white, consider reducing your images to grey-scale.  Not only will your files be smaller but you get a chance to see what the printer will do with them.

When printing to a fancy inkjet "photo printer" you will soon become painfully aware that these printers can easily cost more than a dollar per side to run, if you leave all seven colours blasting away.  If you are printing photos of the grandkids, then maybe cost doesn't matter.  If you are printing your résumé, maybe the investment is worth it.  For everything else, see if you can live with four colours or less.

When printing to an offset printer, get expert advice if you are going to print in colour. Seemingly innocuous choices made before you write your document can double or halve your printing cost: but it takes an expert to know what can be done cheaply and what costs serious money.  Note that each colour you use adds as much again to the base price of the job.  Let's assume your print job is going to cost $15,000 to print single-sided in black.  Then double-sided printing will cost $30,000.  Two colours (black and a single "spot" colour for headings and highlights) will cost $60,000.  If I were you, I would stop there.  Four colours (the minimum for photographs containing flesh-tones) will cost you $120,000.  You know those really attractive plant and animal books, or illustrated histories of the grand masters, that you see on the coffee tables of rich people?  They are often printed in seven colours: $210,000 just to run the job.  Please don't expect that to be the end of it: to get seven colours to perform, you need expensive glazed stock: think about roughly 100 times the paper cost.  Making up seven-colour colour separations is a job for a real top-gun printer with some very serious equipment, and best left to a dedicated pre-press shop.  You wouldn't want to have to run the whole job over, because the expensive fashion model now has orange legs, would you?

Colour Standard

When going out to electronic media, you must use the RGB colour table.  If you do not, very peculiar results are guaranteed.

When printing, at some point you must express your output in CMYK (Cyan, Magenta, Yellow, and Black).  If you need more colours than that, get a pre-press shop to do the separations for you.

Be aware of a bug in the raster filters in various Microsoft products.  Certain versions of Word cannot handle certain combinations of colour in certain formats.  The most likely culprit is the 32-bit colour format that PhotoShop produces.  The formats that tend to give trouble are TIFF and JPEG.  The reason is that Word has a 24-bit colour table, so it is inclined to crash or be unable to display graphics using the 32-bit format.  Then again, Word v.X and earlier on the Macintosh are limited to the old 16-bit colour table: they can't display recent Office pictures from the PC.

The number of "bits" in a colour table refers to the maximum number of bits that can be used to describe the colour and brightness of a pixel.  An eight-bit colour table can describe 256 colours or shades of gray (each individual shade of gray counts as a "colour").  Sixteen bits produces 65,536 colours, 24 bits produces 16,777,216 colours, while 32 bits gives 4,294,967,296 colours.  On Windows, these used to be called "Thousands of Colours", "Millions of Colours" and "High Colour".  On the Mac they still have these names.

Now, this can be a little confusing:  The number of colours in the file says nothing about "which" colours they are.  They can be any colour you wish, provided that you do not have more than the maximum number of colours in a single image.  So you can produce a very satisfactory photograph in 16-bit colour, provided that you do not have two people in the photograph.  What?  Yep:  It takes most of the available number of colours to describe the subtle variations in colour and brightness needed to make flesh tones look realistic.  Provided that you have to deal with only the colours needed for one person's skin, you will get away with it.  If you want a photo of the whole class, go up to 24-bit colour.  Advertising material, where exact colour rendition matters, is often done in 32-bit colour (there are some applications that work in higher colour standards than these, but if you were using one, you would not be needing to read this article!).

Now why on earth would the number of colours matter to you?  Easy:  if the black-and-white photo is 10 kB, the greyscale or 256-colour file is 80 kB.  At 16 bits it's 160 kB.  A 24-bit file is 240 kB, and at 32-bits the file is a massive 320 kB.  The more bits per pixel, the bigger the file.

If you do not need lots of colours, consider using the GIF format, which compresses pictures by reducing the number of colours in the file.

Limiting Factors

Whenever we need to make a decision about a graphics format, the first thing we need to be aware of are the "limiting factors" for our particular document.  There's no point in storing or sending 32-bit graphics to a WAP telephone display: the cell phone will simply remove information or crash, and we will have wasted bandwidth, disk space, and time on colours or details that can never be used.

On the other hand, it is very important not to err too far in the opposite direction: if we do not start with enough information, the result looks really ugly on a high-performance device.

This can get quite complex when you are working on a document that is going to be used in multiple media.  One way to work this out is to look at the limiting factor for each output medium.

First Find Your Upper Limit

Next Find Your Lower Limit

Tips and Tricks

Generally, it is worth reducing the file size of images if you can: your Word document will be easier to handle and less likely to corrupt if you keep it small.  Here are some strategies:

If you are using a vector format, remove information you do not need to print.  The size of a vector illustration is directly related to the detail stored in the file.  If one object obscures most of another, trim the one that can't be seen.

Files coming in from AutoCad contain lines of infinite precision.  Often, they are of zero-width: they are just lines.  They won't print: remove them.  Sometimes you will find AutoCad drawings containing fills: they look like tiny speckles, but in fact they are a complete curve and often quite large. Cut them out.

Often an AutoCad file shows a simple object such as a screw: if you expand the image far enough, you will find that the drawing contains all the information including the screw manufacturer's name.  When printed, the individual parts of the screw are far too small for the printer to print: remove them.

A vector format can contain all the fonts needed to print the diagram and its properties block and the company logo.  There are several megabytes of font information in there.  If you know that the recipients will have suitable fonts on their computers, you can remove the fonts from the diagram (but don't do this if you are going to a Print Bureau, unless you know they have them too!).

If you are using a raster format, use a compressed format if you can.  For more information, see The Formats and Electronic Output further on in this article.

Keep the original file to one side, and use a resampled copy in the document.  Resample the copy to be the ideal resolution and colour depth for the output medium.  For example: measure the size of the picture on your web page, then resample the copy to be that size at 96 dpi.

If you have a document containing lots of screen shots, set your operating system desktop to 256 colours before you capture the screen shots.  The Windows Classic theme is carefully designed to require only 256 colours to display it correctly, and this dramatically reduces the size of the files.  Having taken your screen shot, save it as a PNG and embed the PNG not the bitmap: a PNG is one twentieth of the size of the equivalent bitmap.

When working in JPEG, keep the original image in TIFF and make a JPEG appropriate to each of your planned uses.  For example, you can create one with a compression of 30 per cent, which will work quite well for general office printing, one with a compression of 50 per cent for use on the intraweb and one with a compression of 60 per cent for use on the World Wide Web.

If your document will be edited only by advanced users of Word, consider whether you could use linked graphics.  Removing the pictures from the file and linking them back in can dramatically improve the size of your document and your working speed.  Look up "linked graphics" in the Word Help for more information.

Don't "crop" images in Word.  Word doesn't remove any information, it simply masks the picture so that not all of it is displayed.  Not only does this mean your file is unnecessarily large, but the cropping always seems to move around, creating maintenance work for you.

The Formats

Use rasters for the best compatibility with other applications, and for continuous shading.  Use vectors for the best resolution.


These are the common raster formats.  If you use anything else, only you will be able to open the file.

BMP an oldie but a goodie.  "Bitmap" format, often referred to as Windows Bitmap Format, was one of the first of the raster formats to appear.  It supports only RGB colour and has no compression at all, so the resulting files are huge.  Modern graphics programs enable you to compress BMP format, but be careful: many programs cannot open a compressed BMP.  All the screen captures you take with Alt + PrintScreen are saved to the clipboard in BMP format.  Re-save them from Microsoft Picture Manager (part of the Office Tools) as a different format before using them.

TIFF (Tagged Image File Format). Originally created by Aldus, for use with PageMaker; and now maintained by Adobe, the TIFF format is the industry standard raster format when print quality is of primary importance. It supports both the CMYK and RGB colour models, resolutions as high as you like (2400dpi and higher), colour depths of up to 48-bit, the very efficient, non-lossy, LZW compression algorithm, and it is supported on all major computer platforms (Windows, Mac, Unix, Linux, etc.).  TIFF also supports several other compression formats, such as CCITT 3, which is the format all faxes use.

GIF (Graphics Interchange Format) is one of the first compressed formats to be developed for the Internet.  The format was developed by Unisys in Australia and quickly became an Internet Standard because of the very high degree of compression it achieves.  Graphics program producers these days steer clear of GIF because Unisys still owns the patent and has threatened to charge for its use.  GIF has the ability to encode the image in interlaced format: which appeals to web designers because it enables them to load the image progressively: the reader gets a low-resolution picture immediately which then gradually comes into focus.  GIF has fallen out of favour recently because it can describe a maximum of 256 colours.  If you create your image using web safe colours, consider GIF because it will be one of the most compact formats you can use.  If you are using more than 256 colours, GIF is not for you.

PNG (Portable Network Graphics) is an extensible file format for the lossless, portable, well-compressed storage of raster images. PNG provides a patent-free replacement for GIF and can also replace many common uses of TIFF. Indexed-colour, grayscale, and true colour images are supported, plus an optional alpha channel for transparency. Sample depths range from 1 to 48 bits. PNG achieves higher compression than GIF or TIFF.

JPEG (Joint Photographic Experts Group) is a lossy compressed format developed specifically for storing photographs.  JPEG achieves the greatest compression of any of the common formats, but you need to be keenly aware that it does this by permanently removing information from the picture.  Like its big sister MPEG (for moving pictures) JPEG uses sophisticated mathematics to attempt to remove from an image only the information that the eye can't see.  JPEG output filters have a "quality setting" that enables you to trade off the size of the file versus the amount of quality you are willing to lose.

ICO (the Windows Icon format) is a hybrid raster file that contains the same image in multiple sizes.  You will run across it when customizing applications, creating VBA projects, and creating web pages.


Currently there is no vector format that is utterly safe for use on the web. 

AI is the native format of Adobe Illustrator.  It is actually a form of EPS.  Don't put it in Word documents because not all versions of Word can handle it.  Often you can fix problems with AI by reopening the graphic in Illustrator and saving it again with an RGB or Black-and-White TIFF or a WMF header.  Word, remember, cannot handle CMYK colour in raster formats.

CDR is the native format of CorelDRAW.  Word includes a CDR filter, but it's not a good idea to use it, because the filter cannot cope with the latest CorelDRAW format.

EPS (Encapsulated PostScript) is the publishing industry standard.  It results in the largest file, but has almost no limits in what it can describe.  However, you need to be aware that a reader must have a PostScript printer to be able to print EPS.  And it must be genuine PostScript: "emulations" and "compatible" PostScript is not good enough.  You must also be aware that EPS represents a very tough "final exam" for the printer driver or rendering software at the destination.  While the format itself has almost infinite limits, it is possible to produce a file that will crash the receiving device.

Note that an EPS file can have a "placeable header" in a different graphics format.  The industry standard for the header is TIFF, but high-end graphics packages can also produce WMF headers.  Working in Word, you should produce your illustrations with WMF headers where possible.  When Word prints to a non-postscript printer, it falls back to sending the placeable header when  the printer cannot handle the PostScript information.  If it prints the TIFF it looks utterly disgusting.  If it sends WMF instead, most users cannot pick the difference between WMF and EPS.  This is ideal for office printing: EPS with a WMF header will print or display well on any thing anywhere.

You can produce an EPS file without a placeable header.  If you do, later versions of Word will display a square the size of the image containing two lines of text saying that the image is EPS and will print only on a PostScript Printer.  However, earlier versions of Word will not be able to handle the picture, and neither will some other applications.

EMF (Enhanced Meta File) is an enhancement of WMF, containing 32-bit numbers.  It is the native graphics format for Windows, and has much higher limits than WMF in what it can describe, similar to EPS.  Don't use EMF unless you have to, because it is a bulky format that not all applications can handle.  Macs can’t handle it.  However, it is a life-saver when your image won't render in WMF.

SVG (Scalable Vector Graphics) is the vector format of the future on the web.  Which means don't use it anywhere in the present, unless you have first checked that all your intended recipients can handle it!  SVG is a W3C Recommendation (which is as close as you can get to a standard on the Web).  Sadly, Microsoft products do not produce it.  Instead, Microsoft is attempting to offer its own "standard".

PDF has been with us so long that it has become a de facto standard.  In reality, it is a proprietary Adobe format based on PostScript.  PostScript is a "page layout language": it assumes that the destination is a printer and it provides enough information to tell the printer where to place ink on the paper.  It’s the native graphics format on Mac OS X.

Be aware that PDF was designed and is being enhanced to prevent the recipient from changing or re-using the information they are sent. As far as possible, PDF prevents anything other than viewing and printing.  There is indeed a market for the ability to prevent the user using or changing your material.  However, modern documentation practice requires users to collaborate with each other: to share and re-use information in each other's documents.  PDF is designed to prevent that!

PGML (Portable Graphics Markup Language) is Adobe's answer to SVG.  We don't yet know of any applications that can handle it.  It has capabilities similar to EPS.

VML (Vector Markup Language) is Microsoft's suggested alternative to SVG.  It has the benefit that anyone with Internet Explorer 5 or above and Office 2002 or above can use it.  The drawback is that absolutely no one else presently can.

WMF (Windows Meta File) is the safest vector format of all to use.  Almost every single piece of Windows or Macintosh software can handle WMF.  It was the native graphics language in Windows prior to Windows NT, and is so ubiquitous that everything can use it.  However, it is based on 16-bit numbers, which means its ability to describe colours and complex curves is limited.  If you attempt to save a complex drawing such as an AutoCad file into WMF, it will often "explode", with jagged zigzag lines shooting off the picture as if a graffitist has scribbled on your drawing.  When this happens, try EMF instead.

XML is the way that the entire industry is going.  It is the native format for Office 2007.  SVG and VML are both applications of XML.  The benefit of XML is that instead of encoding "how to print" an object, it encodes "what the object is".  This enables the receiving application to display it most appropriately for its use.  When writing XML, you do not need to know how or where or for what purpose the object will be used.  The exact same picture can appear on a WAP telephone or a 2,400 dpi seven-colour offset printing press.  The receiving device decides how much information it wants and what to do with it. As a simple example, when I write my document I don't need to know whether you will be printing it on letter or A4 paper: I don't need to know whether it will appear in monochrome or colour, and I can look up your local ICC (International Colour Consortium) colour profile to correct the image for your monitor, your printer and your graphics card.

Looking in the crystal ball, SVG will be one of the winners.  SVG is supported by all the industry heavyweights.  PGML will be with us as long as Adobe manages to keep PDF with us.  VML will be around as long as Microsoft keeps supporting it.

In Office 2003, use PDF for vector graphics on the web.  On the corporate intraweb, consider VML if you have a controlled environment with a specified browser (IE 5 or above) and a common desktop environment that includes Office 2002 or above.  For the future, consider moving your operation to SVG as it emerges.

More About Formats

The above definition of a vector graphic is over-simplistic, because all mainstream vector graphics formats can store both vector and bitmapped information, and are therefore sometimes called "metafiles". So you can import a raster into a vector graphics package such as CorelDraw or Illustrator, add your own text or curves, and save the result in a vector graphics format, and both the raster and vector information will be stored in the resulting file.

The above definition is also over-simplistic in another respect; all vector formats allow you to store text either as text or as curves. If you store it as text, the printer doesn't see an equation like that described above; instead, it sees: "The letter 'a' in Font 'y' goes here." Then it looks up the font table (either its own font table or one your printer driver downloaded) for the rest of the information needed to print the character. Your screen will also use the relevant font table in order to display the characters instead of trying to render the curves as curves.

Storing the text in a vector graphic as text is generally a good idea if you can be certain that everyone who may ever need to view or print the graphic will have the relevant font(s) installed (so if the text is Times New Roman or Arial, go for it!), but is a bad idea otherwise.


In order to make the lines look smooth in a raster, they can be antialiased by the raster graphics programs that they are created in. Antialiasing reduces the jaggedness of curves by surrounding the "steps" that the fonts necessarily create with shades of grey or colour. Although this reduces the jagged appearance of the lines, it also makes them fuzzier:

The higher the resolution of the graphic, the less need there is for antialiasing, and the less fuzzy any curves will look when printed. Because computer screens display only 96 dpi, increasing the resolution of a raster above 96 dpi will not improve its screen display.  However, if you intend to print the picture, you should use as many pixels as you can get.  This will dramatically improve print quality, especially if the graphic contains intricate curves (such as small text).  Note that we said "as you can get".  Many graphics programs offer the ability to "resample" an image to a higher resolution.  This normally does nothing other than make your file bigger and fuzzier: it cannot "create" information that is not already in the file.

However, re-sampling is a great idea for reducing the size of an image.  If you have a high-resolution image that you need to use on the web, by all means resample it to a 96 dpi, or your users will never download it.

In the publishing world, bitmapped line art is generally created and printed at between 1200 dpi and 2400 dpi, and at those resolutions most human eyes are incapable of distinguishing a well-created bitmapped curve from a "real" curve. But at the print resolutions more common in business and home environments (such as 600 dpi), bitmapped lines, however well created, will always look a little fuzzy.

Hybrid Formats

By "true" vector, I mean that it is no good using a WMF or EPS that contains a raster of the line art or lettering - that is just a raster format with a vector "wrapper". You need to have the original artwork. This may sound obvious, but vector/metafile formats are so little understood by most companies that getting hold of original artwork is usually a nightmare.

If you ring most companies' marketing departments and ask for their logo they will send you a BMP, GIF or (if you're very lucky) a TIFF. If you specifically ask them for an EPS or WMF, they will save their raster image in EPS or WMF format and send you that!! If you patiently explain to them that a BMP saved in EPS format is still a BMP, and that what you need is the original artwork, they won't have a clue what you mean (and if you make the mistake of saying you need a vector image, they'll get more confused still).

I have found that the only recourse is to ask for the telephone number of their design bureau, and to ring them directly in order to explain that I need the artwork. That is, assuming my clients allow me to make the call at all. Sometimes I have given up in despair and recreated a company's logo from scratch, in Illustrator!

Photographs should, in general, be saved in a raster format; but even images containing photographs may often have line art or lettering superimposed onto the photo; and if so, you need to get hold of the original artwork (which usually means an EPS).

Raster files all have a colour-depth saved within the file; that is, not the colours actually used in the image, but the palette of colours available for use; and the size of this palette has a profound effect on the file size.

To give an example that is easy to reproduce, if I capture my entire screen, paste into a raster editor, and save as uncompressed BMP, the resulting file size is 2.3MB. If I save that BMP as an LZW-compressed TIFF, the file size goes down to 78k. If I reduce the colour depth to 256 colours the file size becomes 48k; and if I reduce it to 16 colours file size reduces to 38k, all with no loss of quality.