Poor export quality of PNG and JPG.

[Designer1]

This problem also existed in Affinity Designer 1. Nothing has been improved in terms of export quality in Affinity Designer 2.03. The outlines of the exported font are pixelated and not smooth enough to use for professional client presentation.

Download this example and look at 100% magnification. For example, at 8, contours are pixelated, as well as many other round letters.

[Designer1]

9 minutes ago, Designer1 said:

This problem also existed in Affinity Designer 1. Nothing has been improved in terms of export quality in Affinity Designer 2.03. The outlines of the exported font are pixelated and not smooth enough to use for professional client presentation.

Download this example and look at 100% magnification. For example, at 8, contours are pixelated, as well as many other round letters.

Here you can compare the same test with CorelDraw. The export result is much better, sharper, smoother!

[myclay]

I tried to match the position of the provided pictures and made a comparison.
The "corel draw " picture has apparently a darker black and might look sharper because of it?

 

 

 

[Designer1]

1 hour ago, myclay said:

I tried to match the position of the provided pictures and made a comparison.
The "corel draw " picture has apparently a darker black and might look sharper because of it?

 

 

 

In both Affinity Designer and Corel Draw, text was designed in CMYK. The same colour profile PSO Coated V3 was used in both programmes. Affinity Designer 2.03 exports a grey-black. CorelDraw exports black as it should be.
However, it plays almost no role in smoothing the letters. CorelDraw has no export functions to choose from (bilinear, bikubisch, lanczos) like Affinity Designer. CorelDraw simply exports in better quality, i.e. not pixelated. The antialias is simply better when exporting.

[Designer1]

35 minutes ago, N.P.M. said:

If afdesigner doesn't meet your standards why not use coreldraw?
 .png doesn't support cmyk and Jpeg isn't suitable for text so what's your point of coming here over and over again pointing this out?
Use coreldraw for your text images.
Again use coreldraw for your textimages.

 

 

I like using Affinity Designer. It's a very good app. However, export quality and antialiasing can be improved.

[oscarlosan]

1 hour ago, Designer1 said:

I like using Affinity Designer. It's a very good app. However, export quality and antialiasing can be improved.

export quality is bad in Designer and Publisher if you compare it to photoshop or illustrator. And the weight much higher than in those applications. Solution: I have to export it as pdf for printing and import it with Illustrator and export to png. Also a screenshot has more quality than the export. Please, affinity, can you improve this? I also think it is urgent.

I do not think it is very accurate to say that we use another application. Sooner or later all users will have the same problem. 

[SrPx]

 

I would have some questions (mostly curiosity)

- Which algorithm  for resampling are you using in the PNG export properties, in Designer? I mean once clicking on the "More" button. Bilineal, Bicubic, Lanczos 3 (and if so, separated or non separated). Spoiler alert, I love Lanczos 3. (apparently there's already Lanczos 8, but in my several brands of 2D and 3D  apps I seem yet to only find Lanczos 3. Is thought by many (I do so) to be superior not just to bilinear, also to bicubic).

 - Where are you viewing the results, to compare them? In a specific image viewer ? Which is it? Or are you seeing the PNgs in their respective viewports (in Corel Draw's and Affinity's viewport). As that could be viewport-specific, often not true to the actual quality saved in the file, and shown in browsers, etc (also, some browsers apply smart scaling and other stuff, so, a bit of extra variants and factors there...), and very different in one app vs the other.

- Myclay might be onto something: barely lighter blacks do produce a less sharp perception, specially in smaller fonts. When coming from a CMYK image, where the black should be formed (to print well and sharp, but this is for the actual  printing machine, not display) only by black tint (100% "K" while c,my,y at zero). This produces a "light black", not perceived as full black.  Now, the conversion of the CMYK profile towards RGB (PNG is always rgb, it does not support CMYK) could be happening towards  a richer black (a rich black is a black formed by the mix of several colors, obtaining a deeper black than a pure cmyk black of only k at 100% and other "inks" at zero) in one app (ie, Corel) than the other. Resulting in a quite lighter "black" in one. Where it happens so, the font is no doubt going to look less sharp, as has less range for everything, and way less contrast to play with. So, I would try to make sure that I am using -as a result- the same level of black exactly, once converted to PNG, as yep it can affect our perception of sharpness.

- Affinity apps do a sort of viewport optimization for performance (Photoshop used to do this for zoom factors other than round numbers, today I dunno) , so, if you are evaluating the exported png sharpness in Affinity viewport, you might be seeing less sharpness than the png will show in image viewers, other graphic apps, or even a web browser.

- I  made a recent test (related to other matters) with PNG and PSD exports, and concluded that the export when using Lanczos 3, and "separated" or "non separated" (can't remember which now) produced a sharper export, but then I realized that when myself applying (to the less sharp one) an unsharp mask with good values (that are typically always the same for low res graphics), the export would look as sharp as the export of any other "standard" 2D application. As if in Affinity, it could be leaving the  unsharp mask pass to your criteria. I would give a chance to play with unsharp mask (Photo's live filter and normal filter, but I believe Designer's Pixel Persona has it too) and then compare the best output you can get of that to the 8pt font export from Corel, and see then what is the conclusion. If the unsharp mask is the solution, as it was in my case, I'd just add it as a live filter to my projects, almost by default, when I'd want very clear and sharp edges.

I am not stating anything. This is just what I would do if I'd find such a problem myself.
 

[Designer1]

11 hours ago, N.P.M. said:

Why so confused?
This post is addressed at @Designer1

 

 

This shows that although the problem is known and has been discussed several times, Affinity Team is not taking care of it. That's a pity.

[Designer1]

8 hours ago, SrPx said:

 

I would have some questions (mostly curiosity)

- Which algorithm  for resampling are you using in the PNG export properties, in Designer? I mean once clicking on the "More" button. Bilineal, Bicubic, Lanczos 3 (and if so, separated or non separated). Spoiler alert, I love Lanczos 3. (apparently there's already Lanczos 8, but in my several brands of 2D and 3D  apps I seem yet to only find Lanczos 3. Is thought by many (I do so) to be superior not just to bilinear, also to bicubic).

 - Where are you viewing the results, to compare them? In a specific image viewer ? Which is it? Or are you seeing the PNgs in their respective viewports (in Corel Draw's and Affinity's viewport). As that could be viewport-specific, often not true to the actual quality saved in the file, and shown in browsers, etc (also, some browsers apply smart scaling and other stuff, so, a bit of extra variants and factors there...), and very different in one app vs the other.

- Myclay might be onto something: barely lighter blacks do produce a less sharp perception, specially in smaller fonts. When coming from a CMYK image, where the black should be formed (to print well and sharp, but this is for the actual  printing machine, not display) only by black tint (100% "K" while c,my,y at zero). This produces a "light black", not perceived as full black.  Now, the conversion of the CMYK profile towards RGB (PNG is always rgb, it does not support CMYK) could be happening towards  a richer black (a rich black is a black formed by the mix of several colors, obtaining a deeper black than a pure cmyk black of only k at 100% and other "inks" at zero) in one app (ie, Corel) than the other. Resulting in a quite lighter "black" in one. Where it happens so, the font is no doubt going to look less sharp, as has less range for everything, and way less contrast to play with. So, I would try to make sure that I am using -as a result- the same level of black exactly, once converted to PNG, as yep it can affect our perception of sharpness.

- Affinity apps do a sort of viewport optimization for performance (Photoshop used to do this for zoom factors other than round numbers, today I dunno) , so, if you are evaluating the exported png sharpness in Affinity viewport, you might be seeing less sharpness than the png will show in image viewers, other graphic apps, or even a web browser.

- I  made a recent test (related to other matters) with PNG and PSD exports, and concluded that the export when using Lanczos 3, and "separated" or "non separated" (can't remember which now) produced a sharper export, but then I realized that when myself applying (to the less sharp one) an unsharp mask with good values (that are typically always the same for low res graphics), the export would look as sharp as the export of any other "standard" 2D application. As if in Affinity, it could be leaving the  unsharp mask pass to your criteria. I would give a chance to play with unsharp mask (Photo's live filter and normal filter, but I believe Designer's Pixel Persona has it too) and then compare the best output you can get of that to the 8pt font export from Corel, and see then what is the conclusion. If the unsharp mask is the solution, as it was in my case, I'd just add it as a live filter to my projects, almost by default, when I'd want very clear and sharp edges.

I am not stating anything. This is just what I would do if I'd find such a problem myself.
 

Compare the export results in Windows Paint.

[myclay]

Comparing is impossible since the given example images aren´t the same size and thus its skewed in favor of Corel draw.
The Affinity example you posted is 1500x1061px
the Corel draw output is 1800x1273px

[Designer1]

1 hour ago, myclay said:

Comparing is impossible since the given example images aren´t the same size and thus its skewed in favor of Corel draw.
The Affinity example you posted is 1500x1061px
the Corel draw output is 1800x1273px

You are free to export and compare a typographic PNG from Afinity and Adobe or Corel. My opinion after many comparisons is clear: Afinity Designer 2.03 exports in a worse quality. It is a pity that Serif has not improved this yet it.

[Designer1]

4 minutes ago, N.P.M. said:

Same size/different dpi from 2 different programs can you tell which is better or worse?

Please provide PNG. Export more round letters in large 12pt or 10pt. Then one can compare.

[Designer1]

9 minutes ago, N.P.M. said:

Same size/different dpi from 2 different programs can you tell which is better or worse?

The 2nd text from above is from Affinity.

[Designer1]

39 minutes ago, N.P.M. said:

And so is the bottom one so it's just your perception.
 

For example, there are special professional procedures to assess the quality of photography, i.e. how well a camera takes photos. It is determined whether a photo is sharp up to the edge and whether colour reproduction, contrast, brightness, etc. are correct. A comparable professional assessment would be necessary to determine the weaknesses of Affinity Designer professional's export quality. This is where Serif is needed to improve the export quality on the basis of objective criteria. 

[Designer1]

11 minutes ago, N.P.M. said:

12 pt Open Sans

Why are you showing a screenshot of Publisher? I am writing about Affinity Designer. Also, I have no idea from which program you exported PNG and why you opened it in Publisher and then took a screenshot of it, which also spoils the result.

[Designer1]

25 minutes ago, N.P.M. said:

It doesn't matter as publisher is the same as designer.
I exported from 2 programs affinity and xara and exported to png with different dpi (1 from each at 300 dpi and 1 at 1200 dpi all 24 bits)
I brought them all into publisher and lined them up to make sure they are at integer positions.
Then I zoomed in at 100% 
As you can see they are equally good or equally as bad.
Now the question for you is, which is which.

XARA is not a professional design software. Compare with Adobe illustrator CC or CorelDraw.
And if you are not a professional designer with a degree and work experience, but a hobby user, discussion makes no sense because you have different requirements.

[Designer1]

23 minutes ago, N.P.M. said:

You know that most of coreldraw was "copied" from xara when it was corelxara?
And what about MSpaint which you embrace so dearly to make your point, that's real software for grandma living in the 90's
Your just looking for excuses, bye now.
We'll see you posting that same question again in about half a year
 

MS Paint is indeed a primitive software. It displays the exported image as it is, without any enhancements, such as the browser.

I hope Serif will improve the export quality. I really have much more to do than report bugs here in the forum.

[SrPx]

Yep, while Windows Paint surely is showing a PNG (of course, never a JPG to evaluate sharpness, due to the compression. But I guess that's implied) without many tricks -and yet I'd only trust a view in 100% zoom, in any viewer... and even so, who knows, after all I have been able to realize in many apps, including Adobe's- I would warn people about browsers, though. In some cases they do "smart scaling", ie, whenever not showing an image 1:1 (100% zoom) because the web code is making it to show smaller or bigger. I tested back in the day (maybe browsers have improved now, but anything happening in real time tends to be lower quality, by nature) that those resamples are in lower quality than even bicubic and certainly lanczos 3 resample (and of course, that many browsers don't have color management, or not "ON" by default, but  that's another issue, tho can affect contrast perception, and so, sharpness perception, which can be happening also here with different conversions in different apps from CMYK to the PNG's "forced" RGB). And I would not be surprised if in some browsers (there are many: Safari, Opera, Chrome, Firefox, Edge, etc) there's some resampling by default even if the web page code isn't telling it to show at a different size. Responsive design anyway (which is everything, today), is re-scaling everything constantly, and the browser, to my knowledge, never gets lanczos 3 resample quality.

Anyway, I would go to the process I mentioned/listed earlier, as I myself believed what was being said here by two or three people in the forums (which could be true for some use cases if not knowing the following) about the image exports from Affinity (all apps, as the export engine I believe it to be the same), and proved (to me) the final situation not to be true (IF knowing what to do), when just setting Lanczos 3 "non-separated" -which in Affinity includes strong sharpen in the process- the export was super sharp... indeed, a bit too much. So I realized the best is Lanczos 3 "Separated". As while it outputs things a bit too smooth, it has enough detail and information to gain sharpness without losing info:  While the smoothing happens later, I tested that if you add an unsharp mask filter previous to export (or live filter over everything, affecting all layers), the result is pretty good. btw, I am a professional graphic designer (decades, sigh), 3D artist and illustrator (many years, had to switch fields often). Well, I am also a hobbyist, too . I only say it so that the solution is taken in consideration, as going that path should end up with good exports. If one is trying to export a PNG direct from vector content... for this trick (and if willing to get a super sharp sample) it's needed to flatten/rasterize all first (and undo later, not saving the source file flat, obviously) , something I used to do in Adobe's apps, anyway. From Designer, best is to click "Edit in Photo" in File menu, apply in Photo an 'unsharp mask' filter, then export as PNG.

Edit : But if for some reason (I wouldn't understand that in a professional environment) you did not get the 3 apps, so, can't click "Edit in Photo" for the Unsharp Mask (and other sharpening filters), only have Designer, then you might want to go with lanczos 3 non-separated, if really want a lot of crispness.

PD: Might be my 50 years old eyes, but, exporting direct from Designer 2.0.3, a design with text exported as PNG Bilinear (even bicubic) seems as good as lanczos 3 non-separated. And in any case -as another professional- I am not seeing anything bad in any of the exports (but I am trying to). Definitely even the most strict client would not complain. Besides most of the times the clients know little about technical matters, deeper matters in composition, color, the concept itself would be more of my priority.

[Designer1]

57 minutes ago, SrPx said:

Yep, while Windows Paint surely is showing a PNG (of course, never a JPG to evaluate sharpness, due to the compression. But I guess that's implied) without many tricks -and yet I'd only trust a view in 100% zoom, in any viewer... and even so, who knows, after all I have been able to realize in many apps, including Adobe's- I would warn people about browsers, though. In some cases they do "smart scaling", ie, whenever not showing an image 1:1 (100% zoom) because the web code is making it to show smaller or bigger. I tested back in the day (maybe browsers have improved now, but anything happening in real time tends to be lower quality, by nature) that those resamples are in lower quality than even bicubic and certainly lanczos 3 resample (and of course, that many browsers don't have color management, or not "ON" by default, but  that's another issue, tho can affect contrast perception, and so, sharpness perception, which can be happening also here with different conversions in different apps from CMYK to the PNG's "forced" RGB). And I would not be surprised if in some browsers (there are many: Safari, Opera, Chrome, Firefox, Edge, etc) there's some resampling by default even if the web page code isn't telling it to show at a different size. Responsive design anyway (which is everything, today), is re-scaling everything constantly, and the browser, to my knowledge, never gets lanczos 3 resample quality.

Anyway, I would go to the process I mentioned/listed earlier, as I myself believed what was being said here by two or three people in the forums (which could be true for some use cases if not knowing the following) about the image exports from Affinity (all apps, as the export engine I believe it to be the same), and proved (to me) the final situation not to be true (IF knowing what to do), when just setting Lanczos 3 "non-separated" -which in Affinity includes strong sharpen in the process- the export was super sharp... indeed, a bit too much. So I realized the best is Lanczos 3 "Separated". As while it outputs things a bit too smooth, it has enough detail and information to gain sharpness without losing info:  While the smoothing happens later, I tested that if you add an unsharp mask filter previous to export (or live filter over everything, affecting all layers), the result is pretty good. btw, I am a professional graphic designer (decades, sigh), 3D artist and illustrator (many years, had to switch fields often). Well, I am also a hobbyist, too . I only say it so that the solution is taken in consideration, as going that path should end up with good exports. If one is trying to export a PNG direct from vector content... for this trick (and if willing to get a super sharp sample) it's needed to flatten/rasterize all first (and undo later, not saving the source file flat, obviously) , something I used to do in Adobe's apps, anyway. From Designer, best is to click "Edit in Photo" in File menu, apply in Photo an 'unsharp mask' filter, then export as PNG.

Edit : But if for some reason (I wouldn't understand that in a professional environment) you did not get the 3 apps, so, can't click "Edit in Photo" for the Unsharp Mask (and other sharpening filters), only have Designer, then you might want to go with lanczos 3 non-separated, if really want a lot of crispness.

Thank you for these tips. However, one can expect Affinity Designer 2.03 to export a text always and without triks, flawlessly. It is clear that PNG and JPG will never be as sharp and smooth as e.g. PDF (vector), but it should be able to reach the export quality of Adobe Illustrator CC. Finally, Affinity Designer 2.03 is an alternative for Illustrator CC. That's why I don't want to settle for inferior quality.

 

[SrPx]

Ouch... SORRY. I really need to sleep more.

I have just realized I understood badly the issue, since the beginning.. It'd be bad to delete all my posts now, for the thread's coherence. You mentioned "pixelated", and not smooth enough, while for some strange process in my brain (prolly too fast reading and being sleep deprived) I was thinking of blurriness, the opposite. If things get too pixelated, my general rule is to export at very large pixel size from the vector app, and then resize on the raster app I think provides me with best pixel output. This typically provides me with best line and font quality for a client sample (the rare cases when I don't just use PDFs for that) That is, I often leave this to batch processing (often with lanczos) in irfanview, XnView, or even just Photo, as from a big image and treating it with raster filters, you have lot of control.

But yep, apologies as I had not understood well the matter, thinking of "poor quality" as blurriness. I mean, I initially read pixelated and that, but seems my brain went to other posts that previously appeared here. Some of what I say (as tricks, never solutions) still applies, but using other filter, not unsharp mask.

[Designer1]

I suspect that not good export quality of PNG and JPG, especially with typography and round graphics is related to a not optimally programmed antialising profile.

[debraspicher]

I've noticed a difference in the anti-aliasing aesthetic between programs and I agree there's a 'blurrier' texture to Affinity's anti-aliasing, but could it be that it is because they are using a very neutral LINEAR (0~100% flat ramp) anti-aliasing profile by default. Affinity does tend to favor more neutral design decisions (for better or worse) which involve less hand-holding. Sometimes it requires more adjustment by the designer. The downside, absolute laypersons really don't know why things look weird and will see it as less professional than other software. (For example, having to be aware we have to pixel align even artboards 🙃).

I suspect that other programs use a "custom" ramp in order to minimize the "gradient-esk" glow look around the font. Because a straight linear line to to bottom, no curves or adjustments, that's really what anti-aliasing does, it's basically a grayscale gradient outline or outer glow FX set to black with a "Normal" blend mode. FWIW, I don't know that I call that pixelation, if it's not a hinting issue. In PS, we were allowed to set "crisp" settings, etc, specifically for type whereas vectors kept the native anti-aliasing. I don't remember us being able to adjust anti-aliasing at all in AI. However, I suspect it uses a custom/optimized ramp. Affinity allows us to customize it down to our choosing, but then it's not the default experience, so a person must know they have to play with the ramp to clarify edges:

Of course, if Affinity does decide to adjust the default profile/ramp, people are going to pitch a fit if they see something in their vectors change and not understand why. So I really do think they should implement a default profile setting IN PREFERENCES for us obsessive folk who need absolute control over our crispy sharp edges and modify the program's experience this way. I may suggest it in feature requests.

[Designer1]

On 1/27/2023 at 2:00 AM, debraspicher said:

I've noticed a difference in the anti-aliasing aesthetic between programs and I agree there's a 'blurrier' texture to Affinity's anti-aliasing, but could it be that it is because they are using a very neutral LINEAR (0~100% flat ramp) anti-aliasing profile by default. Affinity does tend to favor more neutral design decisions (for better or worse) which involve less hand-holding. Sometimes it requires more adjustment by the designer. The downside, absolute laypersons really don't know why things look weird and will see it as less professional than other software. (For example, having to be aware we have to pixel align even artboards 🙃).

I suspect that other programs use a "custom" ramp in order to minimize the "gradient-esk" glow look around the font. Because a straight linear line to to bottom, no curves or adjustments, that's really what anti-aliasing does, it's basically a grayscale gradient outline or outer glow FX set to black with a "Normal" blend mode. FWIW, I don't know that I call that pixelation, if it's not a hinting issue. In PS, we were allowed to set "crisp" settings, etc, specifically for type whereas vectors kept the native anti-aliasing. I don't remember us being able to adjust anti-aliasing at all in AI. However, I suspect it uses a custom/optimized ramp. Affinity allows us to customize it down to our choosing, but then it's not the default experience, so a person must know they have to play with the ramp to clarify edges:

Of course, if Affinity does decide to adjust the default profile/ramp, people are going to pitch a fit if they see something in their vectors change and not understand why. So I really do think they should implement a default profile setting IN PREFERENCES for us obsessive folk who need absolute control over our crispy sharp edges and modify the program's experience this way. I may suggest it in feature requests.

I find the standard antialiasing setting in Affinity Designer 2.04 not optimal. Maybe this is why the round shapes, especially typography, are not exported optimally. This could be the reason. However, I don't know which antialiasing setting would be a better one. I expect Serif to optimise the default antialiasing so that the vector graphics and typography do not appear pixelated when exported to JPG and PNG.

[Designer1]

Why is the export quality of PNG and JPG not improved in Affinity Designer 2.1.0? This was raised by several Affinity Designer users here in the forum a long time ago. It is not an obvious bug, but better export quality is very important. My guess is that it is not the export quality per se, but the antialising setting in Affinity Designer. Images are exported very well by Affinity. The problem is with vector graphics and especially typography. Particularly affected are round shapes, which are exported somewhat uncleanly, somewhat pixelated.

Otherwise, Affinity 2.1.0 is a very good app.

[chillywilly]

I have found that I get much better results if I flatten my artwork before exporting. I'll attach an example. Both versions were made from a 3000 pixel file that was reduced to 400px in the export window and saved as a JPG format using Bicubic and 70% quality.

Dear Affinity Developers — A simple improvement you can make is to automatically flatten artwork before compressing to export formats that don't support vector layers. This would easily improve the resolution of the curves, such as with fonts. The user should not be forced to flatten the .afphoto file before exporting to JPG in order to get the best results. It should be done by the software as it creates the JPG.