upsacling removes the grainy texture from my image

[Nausheen Talha]

whenevr i upsacling my image the grany texture from my image is removed , wat should i do to mainting its original graint texture of image , plus i want t upscle my image

[daniellei4510 · Accepted solution]

Well, for starters, removing grain is not a bad thing. Assets with excessive grain are normally rejected. But that said, wallpaper and background images are a hard sell on Adobe Stock. They are available for free on the internet, and anyone at all familiar with AI can make such assets themselves. 

[Nausheen Talha]

i m not only creating  back grounds , i m trying to create people images , but the biggest problem i am facing it in is , it  creates lot of mistake in m aking hands n eyes , i dont know any perfect site to create people image s

[daniellei4510]

There are none. Some are close, but you will always need to edit the results. Hands, eyes, ears, feet, skin texture, etc. 99.9 of all AI assets need editing. Or being consigned to the trash. Creating AI is easy. Fixing it is hard and time-consuming. 

[Abambo]

i m not only creating  back grounds , i m trying to create people images , but the biggest problem i am facing it in is , it  creates lot of mistake in m aking hands n eyes , i dont know any perfect site to create people image s

By @Nausheen Talha

The keyword here is: manual correction. No generative AI currently produces 100% correct results. And my prediction will be: The day a generative AI produces correct results, generative AI in stock databases will become obsolete because anyone around the corner can produce good assets without any effort.

[daniellei4510]

"And my prediction will be: The day a generative AI produces correct results, generative AI in stock databases will become obsolete because anyone around the corner can produce good assets without any effort."

Maybe. Except that text to image AI programming companies have no interest in creating picture perfect AI specifically for stock. They are, and will continue for some time, profiting from people content on producing robots, celebrities, political entities, fantasy art, etc. Why invest more time in research and development when people are more than willing to spend money producing low-quality, far from perfect results despite its shortcomings that many don't even take note of or care to fix? There also seems to be a "Hey, look! A squirrel!" factor going on right now. Many have stopped developing text to image AI to the point of perfection and turned their attention to video. Before that, there was in-painting, out-painting, and remixing. They figured out how to do hands and feet "close enough," so now let's move on. Yes, it could happen, but as long as people are willing to spend money on creating AI (even when they have no interest in earning an income from doing so), further improvements are not going to happen any time soon.

But that's fine with me. 🙂

[Abambo]

Except that text to image AI programming companies have no interest in creating picture perfect AI specifically for stock. They are, and will continue for some time, profiting from people content on producing robots, celebrities, political entities, fantasy art, etc. Why invest more time in research and development when people are more than willing to spend money producing low-quality, far from perfect results despite its shortcomings that many don't even take note of or care to fix? There also seems to be a "Hey, look! A squirrel!" factor going on right now. Many have stopped developing text to image AI to the point of perfection and turned their attention to video. Before that, there was in-painting, out-painting, and remixing. They figured out how to do hands and feet "close enough," so now let's move on. Yes, it could happen, but as long as people are willing to spend money on creating AI (even when they have no interest in earning an income from doing so), further improvements are not going to happen any time soon.

By @daniellei4510

My experience is, looking at PowerPoint presentations and printed adds, that many people are delighted with low-quality assets. And yes, why getting the last pixel perfect, which is an enormous work to do, when you can make more money with new features.

[RALPH_L]

Upscaling stretches the image. This results in the software trying to feel in the blanks. There is software, such as from Topaz, that use AI to do this. There are also products, such as Photoshop, where you can add grain manually.

[Abambo]

Upscaling is an Art in itself. The upscaling program needs to create information by interpolating neighbouring pixels. Some upscale programs do a good job with that, but somewhere there is a compromise of what can be done. When you upscale noise, you get countless blurry patches. Or the upscaler denoises the picture first, to get a better base to upscale and preserve the important detail. There is no easy solution.

[Jill_C]

Added grain usually looks like digital noise, and that is usually reason for rejection. Designers can easily add grain to the their compositions if that's the look they're going for. 

[Abambo]

Even adding film grain, that looks like real film grain will be a reason for a rejection. All those filters are easily applied by the buyer, but nearly impossible to remove if unwanted.