Magnific Upscale + FREE Alternative!
HTML-код
- Опубликовано: 12 янв 2024
- Magnific.ai is a crazy good Upscale Website. It is especially good at re-imagining Details of a highly upscaled image. I also provide you with three FREE Workflows that you can run yourself on your local computer :)
#### Links from my Video ####
magnific.ai/
Photo Portrait Upscaler: openart.ai/workflows/18UZFMnp...
Landscape re-imagined Upscaler: openart.ai/workflows/oliviosa...
Portrait re-imagined Upscaler: openart.ai/workflows/5l3RcKm6...
#### Join and Support me ####
Buy me a Coffee: www.buymeacoffee.com/oliviotu...
Join my Facebook Group: / theairevolution
Joint my Discord Group: / discord
AI Newsletter: oliviotutorials.podia.com/new...
Support me on Patreon: / sarikas 
Хобби
#### Links from my Video ####
magnific.ai/
Photo Portrait Upscaler: openart.ai/workflows/18UZFMnpHBz41REGv7rx
Landscape re-imagined Upscaler: openart.ai/workflows/oliviosarikas/landscape-re-imagined-upscaler/RPmlWrh5FKULV9MuZnC8
Portrait re-imagined Upscaler: openart.ai/workflows/5l3RcKm6IMo4m8aVcWnF
👋
lot of custom nodes I can't find or can't install, can you make a video helping install all your custom node?
@@ceelentzpotter2110 install comfyui manager and then click on install missing nodes. i explain that in my comfy lessons
getting a ton of error with the landscape one, mainly the controlnet preprocessor node,
here is there error, any way to fix it?
Error occurred when executing AV_ControlNetPreprocessor: [Errno 2] No such file or directory: 'F:\\ComfyUI_windows_portable_nvidia_cu121_or_cpu\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\comfyui_controlnet_aux\\ckpts\\lllyasviel/Annotators\\cache\\models--lllyasviel--Annotators\\blobs\\5ca93762ffd68a29fee1af9d495bf6aab80ae86f08905fb35472a083a4c7a8fa.incomplete'
Edit: issue fixed
Magnific is ungodly priced, there's no way what's going on under the hood warrants those prices. It's literally a glorified Img2Img program. Nothing against Olivio, great content as always.
Exactly. They just use img2img with Ultra Upscale + Tile ControlNet and they are charging the people with these unreal sums of money.
Insulting prices. It’s not even a real upscale, it’s just img2img.
For god sake... it's more than that...
ruclips.net/user/shorts_pswzrkD_h8?si=xOiRl59WkQfTwy9B but more cinematic
No ! It is good 👍🏻
Fake AI upscale rocks !
exactly.. pretty sure it is just a comfyUI nodes running on Enterprise AI cards (main reason they charge so much as they have to buy these $8000+ cards for the server farms
Very interesting workflows! I'll definitely check them out. Thanks, Olivio!
These are great workflows. Thank you Olivio!
My friend thanks for that workflows! You are an artist with comfy
Absolutely BLOWN AWAY! I had to stop and comment.
I’ve watched closely the AI scene for the last couple of years. I was STOPPED dead in my tracks at the midjourney 6 upscale. Whaa-the-faaa… That is NOT trivial what it did! Unreal. Truly unreal!!!
Thanks again for everything Olivio
Thanks for the workflows Olivio😊🎉
It's not an upscaler. It completely changes the images. The very first image you showed became an entirely different face. It's just an image generator that uses reference images.
Agree, not an upscaler at all
It's .(ai) for a reason...
@@TentationAIAnd you are not a tentation ?
it's called enhancer. They have upscaler option as well I think
I have been tempted multiple times to push the join button or support you on Patreon, that's what i think, between this and the comfyui course you're doing great work!
Give in to the temptation! Support your boy. He gives great content and doesn't "pay wall" his flows.
Why don't you then?
Olivio was like ''Isn't this 400 dollar product amazing guys? Look at that, incredible... now let me show you how to do it for free'' 😂😂 the GOAT of comfy
I finally used the landscape upscaler. It was awesome once I was able to use it. My comfyui had issues with 2 missing installs. I had to substitute for other nodes. Once I was able to get it working, I really enjoyed using it. Thanks for making it.
I've been *stung* by creators who promise the world but you end up getting nothing. Olivio is fair-dinkum: all of his vids are solid and he gives away a lot for free. His workflows and suggestions are more than worth the price of a coffee... cheers!
Very impressive, thanks again on your workflows - got me to be much more into comfyui
Congratulations on your videos!
I love your content and learn a lot!
Would you have a tutorial on how to backup and restore our version of stable diffusion? Sometimes we need to format the computer and have to reinstall everything again.
Any tips, recommendations or plugins?
as usual awesome presentation and right to the point cool workflows
Thank you for your efforts and time in consistently providing us with high-quality content. One thing I was wondering, do you think that these paid upscaling services are, in reality, exactly the same as what you offer here, using the same technologies under the hood ?
thanks so much!
the catch-lights in the eyes at 11:19 are so different from each other, is there a way make consistent catch lights or something? I haven't seen a lora to do it.. have you?
Thank you so much for this labor. It is simply spectacular!!!
Is there any way to transfer the given workflow to Stable Diffusion? Or will it not be able to handle it the way ComfyUI does?
This is super insightful, thank you for making and sharing this. I was just wondering if you have any idea how your ComfyUI workflow compares to the built-in upscaler in Fooocus?
Great video. I wanted to clean up and upscale some old photos that have jpeg compression artefacts. Any recommended models to do this jpeg clean up or perhaps a video to help? :) Thanks for all the useful work.
Awesome stuff @Olivio. Any recent updates that allow for ComfyUI and/or A1111 to run well on OSX with xformers enabled for faster render times?
This is great!
I've been cracking my skull trying to get more detail without modifying the image.
Fantastic! I did notice on your woman with the crown 4:45 that it made her eyes different colors and that there is a strange hand embedded in her crown on the left side of the image. That's something that I notice in a lot of AI-generated or enhanced images; looks amazing and then you suddenly notice something strange like that hand or extra fingers or something. Anyway, thanks for another great video!
As an upscaler artist I'm actually going to look forward to using this! This will come in handy for my Anakin Skywalker upscale project!
What is an upscaler artist 😂
@@SW-fh7he somebody who upscales pictures for art 🎨
@@SW-fh7heDon’t know but sounds cool. 😂
Looks very impressive. I'll try it on some of my fave generations and see how it goes
FYI Invoke AI 3.6 has tiled upscale too and mimics magnify AI for free.
It is not free anymore either
@@Greco1114 ??? did they changed it? I have newest version 3.6 and it's free. Also Fooocus has the same upscaler too (you can change slider preset value when activating debug mode in advanced tab)
Hi@@DanielSchweinert Unless you have something different, I check just a while ago (because I read your message actually) and it asked me to register and once I did it took me straight to purchase a subscription.
My shows have never looked better, thanks for sharing!
I bought Magnific for a month. It’s pretty fun as a creative tool, but as a realistic upscaler, it’s not worth it as there are free tools for that, like Upscayl.
Awesome work!
Very interesting.
A question on another subject : do you know an AI solution really of for colorizing black and white photos : everything I see in this domain produces until now just very artificial results.
Can't wait to try.
I think you definitely can achieve something similar to the "infinite zoom" effect in ComfyUI with iterative latent upscaling and maybe some controlnet constraints and injecting some latent noise. It is quite difficult to keep the balance between faithfulness and amount of details, and prevent the model from hallucinating prompt objects everywhere.
I would say maybe it's similar to a latent upscale model with very strict edge and depth controlnet?
Would love to see a ComfyUI flow for upscaling abstract art, I use SD and its Extras to upscale my art but it’s never quite the way I want.
Great video, just found about you and immediately you have a new fan! By the way, if you hold Z, you can zoom in and out with your scroll wheel directly on the website :)
Perfect timing needed this just downloaded a movie.
Thanks bud trying this out tonight.
Thank you Olivio
Gonna upscale like crazy now.
No this is actually amazing 🔥🔥🔥
Excellent job
This was really helpful.
That's IMPRESSIVE !
Your are THE MAN! Thanks.
I've tested both your workflow and Kijai's. I'd say yours is better for anime/cartoony type images and Kijai's is better for realistic.
I haven't found anything better for realistic than the CCSR workflow that Kijai dropped.
I have been looking for somthing like that - thank you for saving 40$ per month:)
@Olivio: ImageScaleDownToSize seems to work. I installed comfyui-art-venture under custom-nodes, but comfyui doesn't seem to find this important node. Any comment?
upscaler + detailer 😍
It seems simple at first but the flow charts at the was a little confusing. Can you go through that in more details at a later date. Thanks!
Price wasn’t so much of a problem, I was ready to jump in until I read: 200 normal upscales & 100 large ones. So it’s limited! Even worst: they don’t even define what those normal and large upscales are prior to subscribing. Their FAQ don’t even explain the specs, resolutions etc. Too bad, the technology was impressive, thanks for this video 👋🏻
Thanks for all your work and content... I was blown away by this Magnific results when I saw it first a couple of days ago. This is exactly what I have been missing for ages, but I can't help thinking that it MUST be possible to replicate this with all the free tools there are available? The plans they offer for this "simple" service are insane imo.
Hi Olivio, where do I upload the upscaler models?
thank you for share your job, but i can use it i have a problem with the node "comfyui-art-venture" who can't be install... so desapointed !
Those 1x upscale models seem new (to me). Have you posted a video about those in the past?
I am getting this error at the kSampler....'NoneType' object has no attribute 'shape'. any thoughts?
Nice one
Do I need to download these upscale files in order to use this workflow?
Hello, I cannot load comfyui-art-venture. Someone below has the same problem. I am getting SyntaxError: Unexpected non-whitespace character after JSON at position 4 (line 1 column 5) warning when I hit Queue Prompt as well. I'm on a PC, Windows 10, Nvidia 3090. Every time I upload one of your templates, I have a ton of conflicts showing up in the manager, is this normal? I just installed everything this week, so nothing is out of date. I've followed all of your videos exactly. I seem to be one of the few that has trouble with this. Please help.
same here
ComfyUI is giving an error: comfyui-art-venture not found. (although cloned into custom-nodes).
Thanks!!
Olivio, I'd love to try upscaling some special old photos from my parents' generation, with the intention of recreating them as accurately as possible - but cleaner and clearer. What app/program would you recommend trying?
Look up Upscayl is open source, free.
Hreat work as everytime
Do the images have to go through Magnific's website or can I do it offline on my PC?
The AI really knows how to improve any photo or video.
hello,
i get the following error:
Error occurred when executing KSampler: mat1 and mat2 shapes cannot be multiplied (77x2048 and 768x320)
i don't know what to do u_u'
Hello Olivio, thanks fro sharing. I have the error import failed with the ultimate sd upscale, please help me!
What program do you do it in?
Hi. where do i get all the models from the workflow?
Just downloaded, the quality blew me away!
Will this work on an entire body that's blurry or only a face?
Why are you using sd 1.5 do to the upscales? Is 1.5 better for that use case or is just due to some controlnets available and the speed?
I'm just used to 1.5. I tried SDXL, but didn't see massivly better results. But I will build a SDXL workflow too
@@OlivioSarikas Yeah, In my experience, not too many improvements using SDXL. I just use it cause I can and should get better results.
But I think that the big difference is the upscalers that you are using (the skin and the other one).
Amazing job, thanks for sharing it with the community
Hi. I got this Error with the Photo Portrait Upscaler : SyntaxError: Unexpected non-whitespace character after JSON at position 4 (line 1 column 5)
Same here! 🙂
I opted for the monthly price just to try it out. It does work great...however there are some pitfalls. You have to make sure if you don't want your original image to change but be upscaled to look more realistic, which is the look I was going for, you have to play with the settings, otherwise your picture could come out looking weird. As an example, I had a 3D render of a woman with her back to the mirror, when I ran the upscaler with it's default settings, her legs in the mirror turned the back of her legs into her knees. Seems it's not smart enough to know how a mirror image is supposed to look, major flaw. If you drop down the scale for creativity and bump up the resemblance a notch up you will get the desired results and more realistic finish. I'm not sure if I will continue to pay $39 a month for this product but it does work well once you get the hang of it.
thanks !!!
hi can anyone explain to me how I can use these workflows locally, what do I need to download and how to set them please. would greatly appreciate it
Hello our friend! :)
Does magnific do full resolution previews?
Hi. A lot of the node in your worflow have red outline. Google says a model or other option in the json file you loaded is not on your PC. But there was no info on how to fix this. I did use the manager on the workflow and eyerything is updated.
Yeah, without the list of custom nodes to download manually, this workflow is makes no use, unfortunately.
Though, video is great and very educative.
I will never pay for any of these paid AI services that are just taking an open source code and charging for it.
now upscale hands (where SD struggle the most) LOL
i bet their pipeline looks something like : interogate image -> split image into areas -> enchance -> upscale.
why i'm sure? cause foocus can into smth similar with inpainting but manually.
Can we do it with a video in input ?
How do you like the Topaz upscaler? Not free but not subscription based.
I tried multiple upscalers, Topaz works best.
hello , i have a problem whit the comfyui-art-venture . i can not install it. someone can give me some advice?
same issue currently
same here, I couldn't found any solution, reinstall-fix- update etc. nothing worked
@@HalukTurken same. I did delete the comfyui-art-venture box, but than got a differen error instead
Same problem here, and I tried everything, is frustating @@metanulski
Any solution?@@HalukTurken
Having tried Magnific, I wasn't impressed by the quality. Maybe its better for photorealism. I used it on my own anime girls and actually got better results natively with the upscalers used in AI generation.
Can"t we achieve this with some workflow in comfyui ?
Magnific is wonderful, it's just too expensive!
The Creator is a very nice guy, maybe he might do something in the future that might change your mind
Your lectures are the best. Thank you always!
The price is ridiculous. The fact that you are giving it the okay is sad.
Bunch of "Web thieves "
Avoid at all costs ☝️
selling out
I am realy confused with this video. I have so many errors trying to get Olivios workflow to run. Did somebody actualy made this work?
You have plenty of ways to upscale pretty everything, the real problem is hallucination. I mean, these upscales work good but they are still not suitable for photo restoration, a human subject will transform a bit and will not be recognizable. And if you say Magnific works the same way and need to tweak it, well this makes it a less viable option than SD.
BTW have you tried inserting the Self attention guidance node? Seems to make some difference to me.
Thanks! Love your content! Just the right amount of info and presentation. I am glad you don't put everything behind a "pay wall".
Thank you and Thank you for your support ❤️❤️❤️
Thank you so much, how to get Promotion code please?
is there a free alternative like for webui a1111
Wow, TOP !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
0:05 the first image was not even an upscale. That was just a change to a picture that looked similar to the original. We are being very loose on the word upscale here.
I recommend watching the video at 1080p or higher.
Watching at 720, it seems to blur the fine details.
Anyone having issue with getting controlnet aux nodes to work?
Anyone getting this error?: Error occurred when executing KSampler: linear(): input and weight.T shapes cannot be multiplied (77x2048 and 768x320) on the workflow Portrait re-imagined Upscale ?
you might lack controlnet models or have a outdated comfyui or nodes. please check if you have selected the correct models and go into comfyui manager and click on "update all" wait for it to finish and restart comfyui
Hi@@OlivioSarikas I checked everything, I even install my ComfyUI from zero. But the same error persist.
@@OlivioSarikas could you please share the links of all the controlnet models you are using in these workflows?
@@fabiotgarcia2 if you google controlnet models download you get the link for all models for controlnet
@@OlivioSarikas My suggestions for you Lesson12 is a tutorial about how to avoid this error. Like show which checkpoint we need to choose for each controlnet. What is compatible.
This is weird, in comfy ui you can do this with upscale with model and a post ksampler.
Creativity is just denoise, hdr is cfg, etc and comfy has a nice image compare also with the rgthree nodes.
I guess good on them for making money from people who don't know.
If they're charging that much they should be paying you to do this video. It wouldn't be so bad if they offered a downloadable app, but come on. They don't even offer a trial version?
Share Maginific latest X post and get free credits. More ppl share it, more free credits we get.
Do you know that every time you make an image, they have to pay for that image generation too? Upscaling and animation is expensive to run.
@@BrianZajac Magnific does not have VC money to burn like Pika Labs etc have. Pika has payed millions of GPU costs so we could do free videos.
The problem is it's too expensive for an image maker. (,not true upscaler)