Need a basic reminder of how to scale images for export to different size files

[tdcarp]

I don't use AI all the time, but thought I understood the best way to scale an AI image for export.  But, following the technique, I'm getting some terrible quality export files.  If someone sees something I'm missing....

I started with a 640x640 pixel artboard.  No particular reason for the size choice.

The artboard contains 3 text entries, and an 8-point compass rose.  The rose was built using the star tool, and some triangles.  It ends up including 2 4-point stars, and 8 triangles used just for coloring.

The text entries are simple 2-3 word phrases.

After building the graphics and text, I selected all objects, and scaled them to 20%.  Created a new artboard, and collapsed it around the graphics (Object>Artboards).

I then export to PNG using the artboards.  With one named "100pct" and the other "20pct", the result of the export is the two files.

Given all the scaling appears to be vector objects, my expectation is the graphics in both files would be high quality.

But, when I place or insert the 20pct PNG image, it's quite grainy.

I actually get acceptable results by inserting the 100pct PNG image and then scaling it to 20%.  But there, I'm changing bit images, so the tolerance won't be much.

What am I doing wrong using the approach of scaling vector images into their own artboards.  Shouldn't I get equally good quality in each of the PNG files?

[Jacob Bugge]

tdcarp,

It is important to have the images in the exact desired final pixel x pixel size, forget about resolution which may actually lead to wrong sizes and hence blurriness. It is easiest and safest to work at the desired size.

To get the same pixel x pixel size, Export at 72 PPI or use the Legacy Save for Web where you can check the image size in the Image Size window and see it even before doing it.

It is also important to have the artwork/Artboard placed fully within integer/whole pixel X and Y values in the Workspace. Otherwise the resulting image will become a bit wider/taller and the extension(s) will be empty and therefore be white.

Here is an overview of general advice that is as sound as it may be old fashioned, and it may also make it easier for you, especially if everything is pure vector artwork:

A very common unsuitable way is to Export to PNG (remember to use PNG24 and use Transparency for artwork to be in front of different backgrounds) with a medium or high resolution, such as 300PPI.

To retain the clean and crisp artwork, a PNG must be created at exactly the pixel x pixel size that it is to be used for (or second best at sizes that are powers of 2 times as large).

For clean and crisp artwork avoid JPEG.

Therefore, the safest way is to create the artwork at the final pixel x pixel size and use a corresponding Artboard, then either use the Legacity Save for Web (where you can look in the Image Size window for size confirmation) or Export at 72PPI. In either case, use the relevant optimization (available with both ways); it is also convenient to have 72PPI in the Effect>Document Raster Effect  Settings.

If you have pure vector artwork, you can relax a bit and have the artwork/Artboard at any size (the Artboard must have the same proportions as the final image), then use the Legacity Save for Web and set either Width or Height in the Image Size and Apply (make sure the other value is also correct).

The Legacy Save for Web may be an old carthorse, but it knows its way home, even if the driver is drunk and sleeping it off in the hay in the back.

[tdcarp]

Thanks Jacob

Assuming I understand correctly, the two approaches are:

Use multiple artboards, each with the original graphics, scaled to the desired output dimensions.  In my case, the scaling is 20% and 15% of the original artwork.  I then shrink the relevant artboard to the dimensions of the artwork.  Each of the artboard contents can be created by Use Artboards during the export.

The other technique, the one you describe, is to have one artboard, and scale when saved for web (I use CS5, so that's File>Save for Web & Devices...).  In the save window, PNG-24 and transparency are selected.  Scaling can be done via the Image Size window.

Assuming this is correct, I understand the appeal of using the File>Save for Web approach (only one artboard).

I've started testing each approach to see if there are any result differences, and have found the scaling sizes (pixel dimensions) to be different.  That doesn't make sense, so I'll repeat the tests and post again.  I also think I'm seeing quality differences between the two techniques, between the two techniques.  Again, that could be testing error.

I'm going to assume that vector scaling, whether via a Transform or via File>Save to Web, will be the same quality.

More to come...

Tom

[Jacob Bugge]

Tom,

Assuming this is correct, I understand the appeal of using the File>Save for Web approach (only one artboard).

I'm going to assume that vector scaling, whether via a Transform or via File>Save to Web, will be the same quality.

Twice indeed.

When you scale to some percentage, I believe you may quite easily end up with non integer sizes/positions.

It is nice to see the Safe for Web & Devices again (and also just Save for Web); I say Legacy Save for Web as default CC speak.

[tdcarp]

The "warning" about non-integer dimensions is very interesting, and a new one for me.

My latest testing is to go back to the original artboards and see what density they were set at.  I'll then want to probe the idea of whole integer.  You mentioned in an earlier post...

To retain the clean and crisp artwork, a PNG must be created at exactly the pixel x pixel size that it is to be used for (or second best at sizes that are powers of 2 times as large).

...and powers of 2 sounds interesting.

In my case, the PNG file isn't resized after being inserted in the document.  Do you mean, save the PNG at (say) double the dimension of the final image, and then resize by 50% after inserting in the document?  Also, are we avoiding any fractions just to keep things clean (i.e. no AI-driven adjustments/rounding)?

Tom

[tdcarp]

I've retested and am getting different dimensions of the images between the Export and the File>Save when inserted in a document.

The actual PNG files between the two techniques are close to the same size (pixel dimension), but the images inserted in a document (e.g. Word), are different.  The difference is about 74%, meaning the File>Save Web objects are about 74% of the Export ones once inserted in a document.

Is this the pixel density (e.g. 72 vs 300) warning you gave?

Since I'm assuming the File>Save for Web method has everything at 72ppi, the implication is I'm making the "mixed density" mistake in AI and/or during the Export.

Although I don't understand how this would be the issue, part of the vector graphics (the compass rose) was originally in a separate AI file.  I ended up copying the final AI graphics to a new AI file, and pasted it in AI format.  The 3 text entries that were added were in an artboard that may have been set to 72ppi, while the compass rose came from an AI file set for 300ppi.

Again, more testing....

[Jacob Bugge · Accepted solution]

Tom,

The actual PNG files between the two techniques are close to the same size (pixel dimension),

They ought to be exactly the same. The Save for Web is accurate, so any difference would be something strange in the Export.

but the images inserted in a document (e.g. Word), are different.  The difference is about 74%, meaning the File>Save Web objects are about 74% of the Export ones once inserted in a document.

As I had understood it (I have never had/used Word), an image is inserted at a certain size, and then you can scale, so I am afraid the reason is unknown to me.

Is this the pixel density (e.g. 72 vs 300) warning you gave?

The PPI issue tends to give itself away, by the proportions: going between 300 and 72 you get 0.4 (72/300) or 4.1667 (300/72).

Since I'm assuming the File>Save for Web method has everything at 72ppi, the implication is I'm making the "mixed density" mistake in AI and/or during the Export.

More accurately, with Save for Web you just work in pixels, but you are right in that it corresponds to 72 PPI, so you are probably right.

The 3 text entries that were added were in an artboard that may have been set to 72ppi, while the compass rose came from an AI file set for 300ppi.

There is no resolution setting in AI (apart from the Effect>Raster Effect Settings), and images just have the (original) pixel x pixel size they were created at, so their PPI will vary (wildly) inversely when scaled (wildly).

When you create a PNG, you have nice and orderly pixels with the right positions and colours.

If you halve the number of pixels (both ways), 4 pixels (2 x 2) turn into 1 which is the average colour of the 4, which has the same colour and (relative) position as if it had been made at half the size from the beginning, so nice and orderly pixels with the right positions and colours. The same applies at quarter and eighth size (multiple halvings).

All other resizings will make a mess where colours get mixed and muddled.

Hence the power of 2.

There is no reason to create a PNG at twice the size needed, but if you need a PNG both at a large size and at half/quarter/and so on size, you can just make the large one and use that for both/all sizes.

Also, are we avoiding any fractions just to keep things clean

If you mean fractions in positions/sizes, they give image borders with pixels made from mixtures of artwork and nothingness.

[tdcarp]

Jacob, thanks for sticking with this.  It's exactly what I'm looking for.  I'm starting my testing from the beginning of the workflow, showing the guidance for each step, and why it's relevant.

The output for the test graphics is print, so I started by specifying the new AI file raster graphics as 300ppi.  Although I have no raster in the image, it appears this setting will have no other affect on the output, and will be there in case raster is added to the graphics later.  This value can be changed via a menu command after the file is created.

There's also a color setting, which I set as CYMK, given the print destination.  [Not sure what the guidance would be, other than use CYMK for print and RGB for web.  Haven't thought through "why", other than the windows in AI used for color settings will show in the selected version.  In the case of the test, if the graphics are displayed, there will be a color shift.]

The next decision is dimension.  The guidance is to make the graphic dimensions the same as the output dimensions.  The only question here is whether the guidance means the artboard dimensions should be set to the output dimensions.  I'm kinda hoping the answer is that the artboard dimensions may not matter, but the graphics dimensions do.  What I ended up doing for testing is to make the artboard about double the size of the graphics.  Before finalizing the graphics, I either reset the artboard to the graphics dimensions, or have the output created using the graphics dimensions.

There are actually 2 files: one is the final graphics, and the other the compass rose.  I use Place to insert the compass rose in the final graphics.  Am I correct that, Placing an AI object in another AI object still maintains the vector scaling?

One last issue that has come up is the rulers.  I am using inches as the measure of the output dimensions (i.e. 2" wide), and when the ruler is set to inches, things are as expected.  I noticed while testing that, if I change the ruler unit to pixels, AI uses 72ppi as the conversion rate.  Is the ppi setting for the ruler something that can be set, or, is the guidance to stay in the units being used for the output dimensions?

More to come.  This is turning out to be a good tread, so part of my intent is to leave context of using the guidance.

Thanks again...

[Jacob Bugge]

Tom,

There's also a color setting, which I set as CYMK, given the print destination.  [Not sure what the guidance would be, other than use CYMK for print and RGB for web.

Certainly CMYK/CYMK (the former abbreviation being more common) for print, RGB for web, and starting in CMYK whenever both are used, very much owing to the difference in gamut.

CMYK color model - Wikipedia ​

RGB color model - Wikipedia

And then there is colour management; you can start on something about it here:

http://help.adobe.com/en_US/creativesuite/cs/using/WS52323996-D045-437d-BD45-04955E987DFB. html

https://helpx.adobe.com/illustrator/using/printing-color-management.html

http://www.computer-darkroom.com/

The last one with its links is a bit wider (and deeper, I believe).

Am I correct that, Placing an AI object in another AI object still maintains the vector scaling?

If you mean from document to document, always use Copy Paste (Ctrl/Cmd+C, then Ctrl/Cmd+F or V or B depending on where you want it): placing gives you the PDF equivalent which is vector/scalable but may very well be lacking in other ways depending on the artwork.

I noticed while testing that, if I change the ruler unit to pixels, AI uses 72ppi as the conversion rate.  Is the ppi setting for the ruler something that can be set, or, is the guidance to stay in the units being used for the output dimensions?

Illy (job description Adobe Illustrator) counts pixels as a(n actual) unit exactly equalling points, 1 pt/px = 1/72 in. So no way round it.

This causes confusion between this and of the general concept of pixels as the (smallest possible) area of a certain colour in a raster image (usually square but it can be taller or wider), the latter having no actual/specific size but changing with the actual size of the whole raster image (the PPI being changed inversely with the actual size).

[tdcarp]

Not sure your assumption about Place is accurate, so to restate.

I have 2 AI files:  One is the graphics for a compass rose (all vector), and the other is to final output graphics.  That file has 3 point-text entries.  The other part of the final is the AI file with the compass rose.  I get it there by using the File>Place, which creates a link to the compass rose AI file.

As I understand it, AI generates a low-res version of the linked image, used during AI sessions.  The comment on some of the web sites is that the "user should be familiar with the Placed image, so low-res display for editing is sufficient".

The question is, if I resize the DISPLAY for the compass rose in the final graphics file, is it functioning like vector or like raster?

Perhaps another way to ask the question is via the guidance to keep the graphics dimensions the same as the final output image.  Since I do a little resizing of the compass rose Place in the final graphics, any resizing, if raster, will affect quality.  If the display of the Place in the final file is Live, meaning the displayed image is being recreated from the linked AI image, it should act like vector.

I realize I haven't got the question to be simple.  I've attached a screen shot of the final AI file with the linked compass rose highlighted.  It's that highlighted area I current resize to scale best with the other parts of the graphic (i.e. the text entries).

I think I understand all the other comments in your last post (including my error in the CMYK acronym).  The nature of the ruler (using points) is fine.  I'll just need to remember to measure the output size in the same units as I design the AI file.

[Jacob Bugge]

Tom,

When you Place vector artwork, it remains vector (in full quality) when resized.

What happens if you create, or just save, your source AI document (the one with the compass rose) with Create PDF compatible file unticked, then Place it in the target AI document (the one with the text)?

Apart from that, what if you have more than the desired part (such as the compass rose) in the source file?

And in this case, you can resize it accurately, but only because the linked file is the same size as the artwork/Bounding Box.

(including my error in the CMYK acronym)

The CYMK abbreviation is no error, just rarer,

https://www.brumleyprinting.com/cymk-and-rbg-printing-explained/

https://acronyms.thefreedictionary.com/CYMK

https://acronyms.thefreedictionary.com/CMYK

[tdcarp]

Hi Jacob

Based on the guidance that the AI graphics should be the same size as the output, I'm going to adjust the compass rose graphics to the exact size it will be in the final graphics AI file.  Right now, I scale the Place area in the target AI file to fit the artboard and fit with the text entries.

Since your answer is the Placed vector graphics file will maintain it's vector characteristics if the AI file with the Place is scaled, it MAY not be necessary to change the approach to do a copy/paste of the compass rose vector data in the final AI file.  Just to see what happens, I'll run a test of scaling the final artwork using both techniques.

Will post back the results, along with a summary of the final guidance for the AI file(s).

Also, it looks like CMYK and CYMK mean the same thing, but convention uses CYMK more often.  For purposes of guidance (use RGB for display-destined output and CYMK for print-destined) the AI files are CYMK and 300ppi (if raster becomes part of the graphics in the future.

Thanks

Tom

[Jacob Bugge]

Tom,

it MAY not be necessary to change the approach to do a copy/paste of the compass rose vector data in the final AI file.

In your compass rose case there is no to change the way.

But I am curious to see your test of this (from post #10) when you have the time,

What happens if you create, or just save, your source AI document (the one with the compass rose) with Create PDF compatible file unticked, then Place it in the target AI document (the one with the text)?

[tdcarp]

Jacob

Ran the test to confirm that a Placed AI file with only vector graphics will scale without distortion if the "target" AI file is scaled.

First, you're right, the Placed vector data scales without distortion.  I'm still going to resize the compass rose graphics to the size on the second AI file (with text) to follow guidance, although, I'm not sure that matters in the case.  To remind you of the flow, I resize the Place to better fit the final AI graphics.  If I change the original AI compass rose graphics to the size on the second AI file, I don't think it will matter re: scaling.

The other thing that popped up in the test was related to save a PDF compatible file option.  For the initial test, I left the PDF compatible option checked.  I then ran the test again with the option unchecked and got this in the Place area in the target AI file.  If I export to a PNG, this text shows in the output file.  I couldn't find very useful documentation of the PDF compatible option on the net.  Best I could tell is that there are a half dozen or so file formats that AI can save.  AI is native.  The documentation seems to imply with the PDF compatible option, the file can be opened by some apps other than AI, and that the native AI file will be smaller.  The test file for the native AI file (vs. PDF compatible) is larger by just a couple KB.

Tom

[Jacob Bugge]

Tom,

The last bit of the test was to convince you that it is impossible to obtain the actual (fullly functional) artwork contained in an AI document when you Place it:

Instead you (only) get the PDF part which is a simple/simplified version, and it is only possible if you save with PDF Compatible File ticked, which actually gives you a double document, namely the AI document and a separate PDF document which may give a rather large file size especially because any linked images are embedded.

In connexion with simple artwork like your compass rose there will be no loss, but more intricate artwork, you will get a simplified and less editable version which may haunt you later.

But there is no benefit from using File>Place rather than Copy Paste, the latter further being quite easy with shortcuts, and you can pick exactly the bits to use rather than get whatever is in the source file maybe requiring deletion of unwanted parts, as mentioned earlier.

The above is the reason (set) behind this suggestion in my post #8:

If you mean from document to document, always use Copy Paste (Ctrl/Cmd+C, then Ctrl/Cmd+F or V or B depending on where you want it): placing gives you the PDF equivalent which is vector/scalable but may very well be lacking in other ways depending on the artwork.

[tdcarp]

Got it.  Well done!!

Thinking through Place, the key appears to be that it's a link, albeit that what's displayed while editing the target AI file is a PDF-type image.  Somewhere when researching Place, I saw that the image displayed in the place-holder in the target AI file will be lower quality than the actual placed image, so I knew that I wasn't seeing the actual linked data.

My assumption is, like with any links, one advantage is that the underlying graphic (e.g. compass rose) can change, and the next time the final graphic is edited, the updated data in the link will be included.

There are probably circumstances where having the link would be useful (e.g. lots of AI images that use the linked file), in my case, the compass rose probably won't be updated, so the advantage of the Place link is much lower.  This is where the guidance is to copy/paste.

Will restart the workflow with a copy/paste.

Tom

[tdcarp]

Jacob

Ran the test of copy/paste vs. Place.  The compass rose is very different, whether during edit of the target AI file, or when that file is exported as a PNG.  Because AI is displaying a PDF-compatible version during the edit of the target AI file, I expect some differences.  However, the differences are dramatic.

These images are: the original AI compass rose, and the copy/paste image (from the target AI file.  When the copy/paste is exported to a PNG, it shows exactly like the graphics in the target AI file.  The size difference is simply from the zoom I used to get the screen image.

The method I used was to select all (compass rose AI file), copy, paste in the target AI file (e.g. with text).

Tom

[Jacob Bugge]

Tom,

Am I understanding it in the right way, that when you create the compass rose in the source file it has the appearance to the left, and when you copy that and paste it into the target file it has the appearance to the right (only with the same size)?

[tdcarp]

Jacob

Yes.

The image on the left is a screen shot taken from the open compass rose AI file.  You can see the guides.

The image on the right is after a copy/paste of the compass rose graphics (Select All in the sending file (i.e. compass rose), copy, and paste the results in the target AI file (i.e. with text).  The image is a screen shot of the graphics in the target file.

Both target AI file versions (Place and copy/paste) were exported to PNGs.  The compass rose differences were also there.

Both AI files should now be set up the same when it comes to PPI, color scheme, etc.  The only difference, other than the contents, are the dimensions.  I don't see that the differences being seen are from scaling.

I'll keep testing....

[tdcarp]

Jacob

The difference appears to come from scaling.

The Place creates a space holder the dimensions of the icon artwork (405x405 px).  To fit the scale of the target AI file, that area is scaled to 32x32 px.  Since the PDF image displayed in the Place area is live linked to the original AI artwork, it scales without any distortion.

The copy/paste results in an area 16,383x16,383 px.  [I suspect the difference between these dimensions and those of the Place space holder has to do with PPI.  I'll investigate later.]  To fit the scale of the target AI file, that area is also scaled to 32x32 px.  The result is the distorted image.

Confirmed the pasted artwork is still vector.

Interesting the pasted vector artwork distorts when scaled.

Will keep testing copy/paste, and also look more closely at the paste graphics before scaling.

[Jacob Bugge]

Tom,

The copy/paste results in an area 16,383x16,383 px.

That is the entire Workspace, so something is funny somewhere. And PPI only applies to raster images/effects.

And what Danny said.

[tdcarp]

The reason the entire artboard was copied seemed to do with whether guides were visible.

However, it still doesn't vector scale even though there's no raster.  Again, the one on the left is a Place, and the one on the right is a copy/paste of the AI vector artwork.  The different final scaling dimensions differences are simply how I did the scale (specify exact dimensions vs. scale%).

I look at the Select All again to see if there's something I'm doing wrong.  It's almost like its stretching vs. scaling.

[Jacob Bugge]

Tom,

I believe the compass rose has a stroke, and you have Scale Strokes and Effects unticked; you can tick it in the Transform palette (flyout) options and get the right appearance.

When you scale down, especially obvious when you scale far down, the original stroke will distort the artwork.

[tdcarp]

Jacob,

[tdcarp]

Jacob