Left Handed Brisket wrote in post #18610519
FB max size is 2048 on the long edge, fwiw.
I have no idea if this makes any sense, but I do it anyway ... posting half size, 1024, has seemed to work well.
Definitely export at 100 percent jpg quality and again, sharpen those suckers. Try what would seem to be an abnormally high "Amount" setting and an abnormally low "Radius" setting. Facebook re-compresses jpgs so giving them the least compressed images possible is key to maintaining detail.
I have read that if you upload png, they will not compress them. Haven't tried it though.
I would seriously consider using a JPEG quality below 100/12 if using Adobe software as your basis for comparison. With Adobe software you only get a total of 13 levels of compression, regardless of the scale used. The issue with using the JPEG file format is that it doesn't store the image data as RGB triplets, instead it works in a similar way to old analogue TV systems where the brightness is saved as one channel, and then two chromiance channels are used to provide the colour information. This means that even at the maximum quality setting you are going to suffer from conversion losses, as the data is transferred from one format to the other. At the maximum quality setting you don't actually get any data compression applied to the image, it's just the format change. I have actually seen JPEG files at maximum quality that were larger in size than the size required to hold the three 8 bit colour channels as individual pixels. If you must have the maximum in quality, and no conversion losses then you would be better off using an 8 bpc TIFF with lossless LZW compression. That though is generally not compatible with most hosting sites.
If you use a setting of 80/10 in other words two levels down from maximum you will get a JPEG file that is between 40 and 60% smaller than the equivalent uncompressed TIFF file. I have yet to be able to see any difference in a displayed image, be that on screen, or as a 16×12 300 ppi print between this level of compression, and the maximum quality. So what about measuring the differences between the images? Well the conversion losses are pretty much identical for both levels of compression, when compared to the original image data. In most cases where there is a difference in value for a pixel it will be limited to only one of the three colour channels, and that difference is mostly between ±1 and ±3 levels, with extreme values at ±5 or ±6. Oddly if you look at the difference between the two levels of JPEG compression the differences will be larger than the differences between each JPEG version and the original. Remember that it is the difference from the original that is important here, not that between different JPEG levels themselves.
I'm sure that FB and the other social media services have a target size for the image files that you upload that they will try to reach when they compress your image. After all they don't want to completely destroy the quality, since that would lead to complaints from users. With the smaller files that you get from the lower quality setting you may get lucky and FB might not recompress those images. If they do need to compress the image though the starting point for the new compression run is going to be basically the same, so you won't suffer any more with artifacts in the image than if they started with the max quality JPEG setting file.
If your goal is to sell prints then you really should be limiting the pixel size of the image severely, along with a large and obvious watermark. The problem is that the average person will, more often than not, be more than happy with an A4/8×12 print from a 1000px long edge digital image, printed on plain paper from the cheapest of inkjet printers. Especially if its "free". This is a case where using something like 640px is actually a good idea, since it will be really hard to get an acceptable print, even for the sort of person already mentioned. On top of that I would set the PPI value to something daft like 600 PPI, since that will in a lot of software then report the image size as 1.07" on the long edge. Again putting off the average user when it comes to trying to print the image themselves. How do I know this, it's simple I have seen people that I know try this with images they have pulled from the web.