background
If you need to make documents, PPTs, or pictures for articles, or make videos in your work. There must be some experience in finding image materials on the Internet.
However, the quality of pictures on the Internet varies. Sometimes I find a picture that I like, but the quality is not good and the resolution is too low.
Some people put up with it, but I can't. Putting a low-resolution blurry picture on the finished product, I think it's better not to put it, it will lower the quality of the whole work.
So what should I do, find a high-resolution picture again?
In this article, I will recommend an artifact that I use in my daily life. A software that can make up low resolution into super high resolution based on AI algorithm.
Even a picture that is too blurry will be reborn immediately with the blessing of this artifact!
See the end of the article for how to obtain the software.
one
Don't go around the corners, get the picture right away.
This software is a client-side software, based on Real-ESRGANthe model, which can process the portraits quickly, especially the animation pictures, and make a good enlargement. Currently, only Windows systems are supported, and MacOs systems are not supported. The software is open source and very green. It does not need to be installed and can be used directly by decompressing it. There is a very good UI interface, the operation is also very simple, there are not too many thresholds, even Xiaobai can use it!
It looks like this:
If you want to enlarge the resolution losslessly, you only need to drag in the picture, you can drag in multiple.
I found 2 low-resolution cartoon pictures, the quality of the pictures is very worrying, how worrying, let's see for yourself:
The original information of these 2 pictures is as follows:
The image size of Conan and Haibara Ai is 358*269, and the image size is 30k.
The size of this picture of One Piece is 465*245, and the picture size is 49k.
Drag into the software at the same time, click to start the task
Since AI operations require your GPU, the length of time depends on your GPU performance.
My graphics card is RX 5500XT, the AI enhancement of 2 pictures is not very time-consuming, it took about 6 seconds.
The comparison between the original image and the enhanced rendering directly:
Look at this effect, isn't it very wow!
Data information after 2 pictures are enhanced:
The image size of Conan and Haibara Ai is 1432*1076 and the image size is 2.3MB.
The size of this picture of One Piece is 1860*980, and the size of the picture is 3.3MB.
After calculation, the picture quality has been enhanced by 16 times!
two
Anime pictures are enhanced so well, what about the characters and landscapes. I also ran a test on people and landscape images next.
I found a low-resolution photo of a beautiful lady from the Internet and a landscape photo of Italy, among which there are quite a lot of landscape elements.
The same process takes more than 6 seconds. Directly put the effect comparison chart:
As you can see, the effect is also very good. Looking at the data, it is also enhanced by 16 times. The file size is also around 3-4MB.
three
Next, I ingeniously took the enhanced landscape photos and continued to drag them in for enhancement. See if you can make dolls.
Whether it can be enhanced to the original 16*16=256 times.
If the picture is bigger, the AI algorithm will take a little longer to calculate. It took about 39 seconds to get the result, which is a little shorter than I thought.
The effect diagram after the nesting doll is as follows:
I don't know if you can see the difference.
In this way, the data of the photo after the nesting doll has reached 45.1MB, and the size has reached 5888*4304.
Although I think the picture is very delicate, but if you look closely, you will feel that it is overkill.
Four
Damn me, I still want to do a second nesting of dolls, drag the 45M file in, can I get it out?
The result was set for half an hour, still no result. I'm not impatient anymore. Cancel directly.
finally
This software is an open source software on Github, and the author released a green version that can run directly on windows.
The open source software address is: https://github.com/X-Lucifer/AI-Lossless-Zoomer
The software is still a bit big. In order to facilitate everyone to download it on the domestic network, I put this software on the gitee disk, and everyone can download it by themselves.
Reply in [Yuanren Tribe] aito get the download link.
I am Bo Saidong, an open source author and content blogger, who loves life and sharing. If you are interested in my content, please retweet, like, and follow me.
