I just wanted to get some thoughts about render size. I'm rending an image in Poser using Superfly that will be printed as a 24x36 poster. What do you think would be a good render size to get a nice clean image but not take a full day to render?

Thanks

My renders take a full day even at a 10th of that size. I don't think you'll get a clean render that big in less than a week.

Suppose you want 150 pixels per inch (you really want 300 PPI for a sharp image but let's go with 150) 24 * 150 is 3600 36 * 150 is 5400 So your image is 3600 by 5400 pixels. A total of 19,440,000 pixels.

Now some images clean up faster than others. A reasonable guess is somewhere between 500 and 5000 samples per pixel.

So you may need 19.44 million * 5000 samples which is 97.2 billion samples.

I don't know what your hardware can do. Suppose it can do 1 million samples per second. That would come out to 27 hours.

My hardware is way slower than that.

If you're intending this for a print, I'm assuming you're going to do some post work in Photoshop or another image editor. I would just break the image up into sections, render in four or more parts and just combine the rendered sections in post if you want to break it up over a few days. My workflow also consists of rendering the image @300dpi but 25 - 50% larger than what my intended print size would be, shrink it down in photoshop add a very small amount of sharpening and have a much sharper and cleaner image than the original. You can also get away with slightly less pixels samples when you shrink down the image, just like a high ISO image will look cleaner at a reduced size, the same happens with rendered images.

Is this a timed work?

bagginsbill posted at 7:49PM Wed, 11 December 2019 - #4373043

Suppose you want 150 pixels per inch (you really want 300 PPI for a sharp image but let's go with 150) 24 * 150 is 3600 36 * 150 is 5400 So your image is 3600 by 5400 pixels. A total of 19,440,000 pixels.

Now some images clean up faster than others. A reasonable guess is somewhere between 500 and 5000 samples per pixel.

So you may need 19.44 million * 5000 samples which is 97.2 billion samples.

I don't know what your hardware can do. Suppose it can do 1 million samples per second. That would come out to 27 hours.

My hardware is way slower than that.

it really matters on how the scene is lit, if there is a lot of bounce from the environment, dynamic range of the HDRI if one is used, complexity of the shaders. You can get away with less samples if you use a low res jpg instead of an HDRI for enviro lighting. Small room models (main figure inside a room) will need huge numbers of samples to deal with all the bounce. The shaders can have a huge impact on render speed. I have hair models that HAVE to have the shaders replaced because they slow down the render even on tiles that don't have hair on them.

So the question is what does the scene look like and how is it lit. With this info we could suggest ways of speeding up the render time.

Another thought is to go with a lower quality and investigate the use of an AI "denoiser" to clean up the image post render. Never tried one but they seem to attract interest. If using GPU avoid render to screen with progressive mode as the CUDA performance can drop off as the image size exceeds screen size. Render to queue should solve this - at least it did with the tests I've done.

It all depends on what you want with the print. If it is just to hang on the wall like a painting high detail all over will only distract from the important parts. Select your detailed/sharp areas carefully so they stand out from the background. That is how a traditional painting is made. One starts wide brush and then goes down to improve selected areas with finer strokes.

In Poser terms you would probably do an overall render in a moderate resolution and then in post process patch with partial renders in higher resolution where you think necessary, and use transparency and blur to match.

You could even consider to leave out time consuming details (hair) in the global scene render if you know you will patch the area later anyway.

It may also depend on how you are intending to have this printed. I have a banner of my character which is 5 foot by 3 foot or 60 x 36 inches which was printed by Photobox as I am in the UK. The maximum size that Photobox will allow for upload is 100mb so that stipulated my upper limit for the .jpg file I used. I cannot remember the size I used in Poser but I do remember that the render was close to 20 hours.

I was pleased with the end result, close up you can see some noise in the render but then the cloth used for the banner also has texture so this masks the grain to some degree. From anything over a foot away no grain or texture can really be seen and, as this is virtually a life sized figure that seems to be the closet most viewers get.

I also have a picture which is 30 X 20 inches and is on a special plastic in which the render is applied straight to the surface. There is no particular texture to the plastic so any grain/noise would be clear. Again the render took about 20 hours but as the size was smaller there is less noise. Again I was pleased with the result which stands up to close inspection, and it is clear the skin texture was for more important than any other factor.

This was the picture and the file supplied to Photobox was 7016 X 4960 pixels (8.770 X 6.200 inches at 800 pixels / inch) and 20MB.

Since I do this kind of printing for a living, let me throw in a few thoughts.

The 300ppi standard is based on really old requirements for photo-mechanical reproduction technology, where a 150 line screen was very common. A lot of the technology used for digital printing uses the 300ppi standard as the basis for its printing resolution, which can be 600dpi, 1200dpi, 2400dpi etc. Note the difference between pixels per inch (ppi) and dots per inch (dpi). Do not assume that you have to make an image (PIXELs per inch) match the print resolution (DOTs per inch).

A poster of this size is likely to be printed on an inkjet system, and it is highly unlikely that a regular dot pattern would be used for screening: it's almost certain to use 'stochastic' or 'frequency modulation' (FM) screening. Traditional screens rely on Amplitude Modulation (AM). AM means the dots get bigger to deliver more of a particular colour. FM means that there are more dots of a particular colour where required.

About 25 years ago I was involved in conducting research into how image resolution impacted on perceived printed results, and long story short, for most images the difference between the same image at 200ppi and 300ppi, printed at the same size using the same paper and ink (actually printed on the same page) was, if noticeable at all, negligible.

Images rendered out from 3D apps have not been softened by lenses, scanners, etc. so should not need artificial sharpening to 'improve' them.

A poster this size will be viewed mostly from a distance in excess of 1 metre and will not be scrutinised as closely as a print held in the hand.

Many of the more recent RIPs and print control systems have built-in upscaling engines which can, depending on the image, produce astonishing results even from poor quality originals. I did a job this week where a 900pixel x 600pixel image was printed at A3 (12in x 16.5in) and looked, well, not great, but by no means terrible.

With all that in mind, I'd be inclined to render out to 100ppi equivalent (2400 x 3600 pixels); keep the image in RGB, DO NOT convert to CMYK; and DO NOT apply any sharpening.

cspear posted at 1:05PM Fri, 13 December 2019 - #4373084

Since I do this kind of printing for a living, let me throw in a few thoughts.

The 300ppi standard is based on really old requirements for photo-mechanical reproduction technology, where a 150 line screen was very common. A lot of the technology used for digital printing uses the 300ppi standard as the basis for its printing resolution, which can be 600dpi, 1200dpi, 2400dpi etc. Note the difference between pixels per inch (ppi) and dots per inch (dpi). Do not assume that you have to make an image (PIXELs per inch) match the print resolution (DOTs per inch).

A poster of this size is likely to be printed on an inkjet system, and it is highly unlikely that a regular dot pattern would be used for screening: it's almost certain to use 'stochastic' or 'frequency modulation' (FM) screening. Traditional screens rely on Amplitude Modulation (AM). AM means the dots get bigger to deliver more of a particular colour. FM means that there are more dots of a particular colour where required.

About 25 years ago I was involved in conducting research into how image resolution impacted on perceived printed results, and long story short, for most images the difference between the same image at 200ppi and 300ppi, printed at the same size using the same paper and ink (actually printed on the same page) was, if noticeable at all, negligible.

Images rendered out from 3D apps have not been softened by lenses, scanners, etc. so should not need artificial sharpening to 'improve' them.

A poster this size will be viewed mostly from a distance in excess of 1 metre and will not be scrutinised as closely as a print held in the hand.

Many of the more recent RIPs and print control systems have built-in upscaling engines which can, depending on the image, produce astonishing results even from poor quality originals. I did a job this week where a 900pixel x 600pixel image was printed at A3 (12in x 16.5in) and looked, well, not great, but by no means terrible.

With all that in mind, I'd be inclined to render out to 100ppi equivalent (2400 x 3600 pixels); keep the image in RGB, DO NOT convert to CMYK; and DO NOT apply any sharpening.

Thank you for the information you have provided, I have often wondered just how the printing was done in the modern day. I have been really pleased with all the printing I have had done and I tend to create a photo book of my renders each year. In the last few years I have created lay flat A3 photo books that allow a panoramic double page spread which can be stunning (as long as you remember to keep anything important away from the join in the centre). The site I use for printing will throw up a warning if the file I have uploaded would not produce the quality they are used to providing. It seems to work as I have yet to receive anything I am not happy with.

cspear posted at 10:34PM Wed, 25 December 2019 - #4373084

Since I do this kind of printing for a living, let me throw in a few thoughts.

The 300ppi standard is based on really old requirements for photo-mechanical reproduction technology, where a 150 line screen was very common. A lot of the technology used for digital printing uses the 300ppi standard as the basis for its printing resolution, which can be 600dpi, 1200dpi, 2400dpi etc. Note the difference between pixels per inch (ppi) and dots per inch (dpi). Do not assume that you have to make an image (PIXELs per inch) match the print resolution (DOTs per inch).

A poster of this size is likely to be printed on an inkjet system, and it is highly unlikely that a regular dot pattern would be used for screening: it's almost certain to use 'stochastic' or 'frequency modulation' (FM) screening. Traditional screens rely on Amplitude Modulation (AM). AM means the dots get bigger to deliver more of a particular colour. FM means that there are more dots of a particular colour where required.

About 25 years ago I was involved in conducting research into how image resolution impacted on perceived printed results, and long story short, for most images the difference between the same image at 200ppi and 300ppi, printed at the same size using the same paper and ink (actually printed on the same page) was, if noticeable at all, negligible.

Images rendered out from 3D apps have not been softened by lenses, scanners, etc. so should not need artificial sharpening to 'improve' them.

A poster this size will be viewed mostly from a distance in excess of 1 metre and will not be scrutinised as closely as a print held in the hand.

Many of the more recent RIPs and print control systems have built-in upscaling engines which can, depending on the image, produce astonishing results even from poor quality originals. I did a job this week where a 900pixel x 600pixel image was printed at A3 (12in x 16.5in) and looked, well, not great, but by no means terrible.

With all that in mind, I'd be inclined to render out to 100ppi equivalent (2400 x 3600 pixels); keep the image in RGB, DO NOT convert to CMYK; and DO NOT apply any sharpening.

Wow. That is just invaluable information.

Just to add, Blender 2.81 has the Intel denoiser as a filter in its compositor. Which means you can use that filter on image files, like, say, Poser renders. Denoising has made my life considerably easier from the start, and I've seen comparisons show the Intel denoiser as better than both the nVidia denoiser and the Cycles one. Also, it doesn't need specific hardware or software, which the nVidia one does. That said, all of those denoisers are specifically good at raytraced noise. For instance, the Intel denoiser doesn't work on noisy digital photos or Eevee renders.