Enhance & Upscale AI Art For Printing - FREE Tools 🤖🎨
ฝัง
- เผยแพร่เมื่อ 21 พ.ค. 2024
- You must have noticed that AI Art generated has low resolution and printing, or even using it as a web asset, is not very viable. In this video, I compare Midjourney's different upscaling options, and enhance upscaled images with online upscalers and ChaiNNEr, an open source python library that offers a very sophisticated solution with neural networks to upscale your images. With ChaiNNer your images will have enough resolution for framing and printing - T-Shirts, bookmarks, book covers, mugs, etc. Take your AI art to the next level!
Img Enlarger AI - imglarger.com/
PixelCut - www.pixelcut.ai/tools/image-u...
ChaiNNer - github.com/chaiNNer-org/chaiNNer
Model Database - openmodeldb.info/
NCNN Models - github.com/nihui/realsr-ncnn-... & github.com/nihui/srmd-ncnn-vu...
Timestamps:
00:00 - Intro
00:30 - Understanding Midjourney's Aspect Ratios / Upscalers
03:52 - Upscaling Tools
04:17 - Understanding PPI x DPI
05:05 - Comparing ImgEnlarger AI x PixelCut
06:47 - ChaiNNer
11:17 - Conclusion
JOIN US ON DISCORD 👾
/ discord
Oh, you want to buy me a coffee? How sweet! ❤️ ☕
ko-fi.com/bellacapilla
SUBSCRIBE - Help to spread the love! ❤️️
/ @bellacapilla
NOW ON RUMBLE 😎
rumble.com/c/c-2389300
FOLLOW ME
/ bellacapilla
/
/ bellacapilla
CHECK IT OUT
www.bella-capilla.com - วิทยาศาสตร์และเทคโนโลยี
Thank you so much for the great tutorial. Ones again you presented an awesome tool! Will definitely explore more of the ChaiNNer and how it can be used for my workflow.
amazingly detailed and very very helpful in my search for image enlarging!! thank you so much. subbed for sure!!
What a powerful free tool that ChaiNNer, for that I subscribed ;)
I was looking for something like this, thank you so much
Very helpful! I personally use Nero Image Upscaler to enhance my AI art crom Midjourney. These two tools are really handy.
Thank You!! Very helpful tutorial.
Wow amazing work!!!
excellent tips, thank you 🙏🏻
Thank you for uploading this video
Thanks for an informative video!
You do a very good job, keep it up. One question please, I have not seen how you prepare the color. Do you work with the adobe rgb space and then when exporting you pass it to cmyk? Or do you directly work the file with the cmyk space? midjourney images if viewed with cmyk, change a bit, especially more magenta. Thank you, new subscriber!
Thanks so much for your expertise.
Excellent video. Thank you. ChaiNNer looks a bit complicated, but I'll give it a go.
YOU ARE A GENIUS! Thank you for this tutorial. I have been looking for a solution for a while. I tried LargeIMG, Lightroom, and Photoshop, but this node-based way gives me way better control and way better results. I will be buying you a coffee. Thanks!
I'm sooo happy to hear I was able to help! 😊Thank you so much for the kind words and for the coffee! ❤ It really means a lot to me! 🤗
brilliant! thank you
I'd not heard of this before so thank you - and as you say, if you remember to set up using the NCNN nodes at actually fast on AMD - thank you :)
Thank you it helps!
That,s great shared. Thank
Thank you!
Usted es una maestra de maestras!!! /// You are a masters of masters!!!
😀
Good presentation. 1 sub here ;)
thanks for mentioning that changing the ppi does not affect the number of pixels. many people misunderstand that
I think this year we will see lots of amazing AI tools that will make our lives easier. Those who can use these AI tools smartly will be the gainers. Recently I created amazing images with Blue Willow
It's great to hear that you're enjoying your experience with Blue Willow
Thank you for saying how long it took. I watched another video and no mention that it took a whopping ten minutes so I thought it just was't working.
I had forgotten to ask you if windows 10 is taken into account for this model of graphics card ? Thanks !
Cool 😎
Hello and thank you very much for your very well explained video. I tried this software twice, and it crashed my computer. It could be the graphics card, maybe. Could you tell me if the GTX 1630 model, 1815 MHz GPU is enough to prevent the computer from freezing? Thanks in advance .
Onto the upscaling of my generated doge adventures collection!
Sorry, the detailed results of the images that have been upscaled using ChaiNNEr have what DPI? because I only saw the pixels. Do you need to up the DPI again in Photoshop?
Gigapixel AI - cool tool
Do you reach 300 Dpi after upscaling ?
Thanks
I'm trying to install the neural network for chainner but it says that torch, torchvision, facexlib and einops are missing. What should I do? It's not really advancing in the loading as well.
hmmm... not sure. There are no issues like that reported. Are you on windows? did you follow what I did?
@@BellaCapilla Of course. I ran through everything but pytorch wouldn't install.
How do make image transparent using chainner
Upscaled a batch of 25 images from 512 to 4k in minutes with my Ryzen 9 7950 and RTX 4090.
Amazing! 😍 If you separate your images by style in specific folders, you can use the node Iteration to just loop through them all, one by one. But then they will all be using the same model, so keep that in mind :) They should be similar style-wise.
I have Topaz Gigapixel AI that can upscale 6 times, yet it's still not high enough for print. Would you recommend using ChaiNNer after or before the Gigapixel upscale, and why in that order?
I'm not familiar with Topaz, but I would be cautious when using two software. Some loss of detail might be unavoidable and different software are going to affect your image in unpredictable ways. For example, one software (which runs with a certain ML model "A" results in less reflections) and another, running ML model "B" doesn't handle shadows well... your result will have less reflection aaaand shadow issues. That's just a speculation, but in short, 2 software = 2x loss of details since each works on an image in a certain way.
sorry for the late reply, btw
@@BellaCapilla no worries, and thanks for replying. Have a wonderful day.
5:49 The ImageEnlarger one has 5MB because it is JPG
True :) and therefore, less quality than svg for example. :)
I'm using chainner my RAM is 32GB I'm using a single model but it keeps running forever how much time does it take to upscale usually?
A few minutes for 1 upscale job. Try stopping and restarting when it takes tooo long :) seems to solve the issue
is there somehwer to get NCNN models? doesnt seem to take PTH
I updated the description with two links I was able to find for NCNN :) you basically need two files, a .bin and a .param
@@BellaCapilla thank you ! I saw that pytorch has a path to NCNN conversion node . Would that work just as good as going with .bin and .param ?
definitely! Give it a try :)
Any idea why i am getting "Failed to Fetch" error message on Mac?
Well, that's for sure not enough information, is it? :D If you mean ChaiNNer, did you read their docs about iOS use? If you mean one of the online tools, their server might be offline...
Followed the instructions for chaiNNer correctly, but the first upscale job started 1 hour and 25 minutes ago and hasn't finished yet.
wow :o how many times you iterated the upscaling? Do you have a GPU?
@@BellaCapilla Not sure what you mean. I linked up everything as you instructed, used Realistic Rescaler, then pressed the start button. My laptop has an Intel processor, 1.8GHz, RAM 8 GB, NVIDIA GeForce MX110. The upscale didn't complete.
@@BellaCapilla Further to my last message, I stopped the processing, then tried again. This time it completed. It took about 5 minutes. 👍
@@BrianAbbottGeordie wonderful! 😍
my pytorch didnt show up after i downloaded it...on the nodes section
Amazing! Thank you for the sub! 😍 Are you certain? It normally appears on the bottom of the left side bar. Try reinstalling it, or restarting pc 🙂 If you have a git account, you can always ask for help here as well: github.com/chaiNNer-org/chaiNNer/issues?q=
can you plz help me correct my certificate which is burned
Was kinda disappointed you didn't use an image of a potato for the upscaling. 😂
hahaha! that is a great idea! 🤣 I'll remember that for the next videos 🍠
This worked really well. My only issue is that it took 3 hours to do 4 images. I only wanted to double the pixels, so I just chose the 2048px size instead of auto. But it still took forever. I have a high-powered PC, so I was quite surprised. Running a 3080Ti with lots of VRam and 64gb of Ram. I was expecting it to be a lot quicker. I'm going to try it again but this time I'll try to use a small file size and have it converted into a larger file for printing purposes.
Interesting. That is really not usual :O If you notice it's taking longer than it should, try restarting the process, might help :)
@@BellaCapilla Yeah I knew something was wrong. I updated my driver for my GPU, also updated my computer. Then I reset my PC and made sure I didn't have anything running in the background. Ran 1 small PNG file of a 1024x1024 and enlarged it to 4800x4800. Took almost 2 minutes and holy cow. This thing is magic. It looks so good. I'm so glad I came across this video. This beats every alternative I've tried, and I tried a lot. Including photoshop and adobe illustrator which was the previous best. But this blows that out of the water. Thank you so much!
@@zeaKz1 I'm glad you managed! :D Super happy it worked for youuu! Enjoy ♥
ich will ja nicht klauen ich suche ob man seine Orginale von einer ai gemacht kann Bilder die ich selber gemacht habe mit meiner Kamera ich finde nur nur irgendwie nur quatsch.
Not sure what you said there. I understand German but your sentence is very confusing... Anyway... You can find models in the models library for photography upscaling to use with ChaiNNer
@@BellaCapilla He's speaking gibberish, even as a native German speaker I don't exactly understand what he means but I believe he's actually looking for stable diffusion, so he can turn his own images into stylized art without having to steal work from others...
i know so simple way and free
Sorry but you said it was very simple with nothing to download, but it is quite the opposite... This is more complicated and there is more to download than you said
What's the most funny is that you said "You don't need to dowload Python or anything, but yes : actually must download python, not to download the software but right after, to make it work properly
I will uninstall it, I cannot waste my time to download everything and see how it all works, I need to create and upload my work... I have some upscalers already but I am always looking for the best one I can use, but I want something made for human brains and not talking to AIs, meaning something easy to understand and to use
By the way, the address is not good anymore, everything has been removed from this one except some non-functionnal elements, and in the new address there is no model list
The link you said doesn't work anymore says they moved to a new one, which indeed looks much better and it's so much easier to choose a model now! (thank you, I've updated the link in description) - openmodeldb.info/ ... I don't understand when you say that everything was removed? The models are all there...........
Now, when I mention you don't have to download Python, you actually don't, indeed. The software downloads it for you in the specific version it needs to run properly. And you DON'T need any programming skills. This tool uses Machine Learning (that's WHY IT USES PYTHON) to upscale your images and not AI. You don't "talk" to AIs, and it's certainly not that complicated. If that's too complicated for you, maybe try some of the free to use browser tools, which only require an upload, but they aren't nearly as powerful as the ML tool. Out of 25k viewers, you're the first saying that's too complicated. Maybe you just didn't follow the video properly and tried to skip sections, missing important information.
In any case, I hope that, eventually, you find the right and easy tool for you that attends your needs and thank you for dropping by.
Your voice could cut glass please use a mike baffle as with phone/buds it gives me head ache thanks.
How rude, her voice is fine. Use cotton wool in your ears.