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
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!
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.
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.
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
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.. 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
@@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.
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
Would upscaling with Magnific before uploading to Runway help to reduce Gen-2's morphing problems with image-to-video? I'm sure it would be noticeable with Luma but Luma is way too expensive as it is without the added Magnific cost.
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'm learning to focus on upscale, I want to ask, sorry, maybe it's a bit difficult for me to understand. My question is, can this upscale directly enlarge an image without having to add other programs/notes (such as Comfuyi's simple upscale), because what I noticed in the video only changed some of the image details rather than the core of the upscale itself. Maybe there is a setting for how many times we can upgrade, like 2x / 4x in the notes, if so, that would be very helpful, once it works, it's done!
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?
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.
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 ?
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.
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!
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.
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'
tks for the workflows, are you aware that magnific devide the image in smaller chunks and upscale them apart to later ensamble again?, whats is that technique name?
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!
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?
8:55 is better than 2:21, because it's more faithful to the original. For example, it doesn't add those sharp, unnecessary laugh lines that 2:21 does. The laugh lines make her appear older and thinner, imo, and alters her mood slightly.
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.
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.
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.
I get an error saying Conflicted Nodes: ColorCorrect [ComfyUI-post-processing-nodes], ColorBlend [stability-ComfyUI-nodes], SDXLPromptStyler [ComfyUI-Eagle-PNGInfo], SDXLPromptStyler [sdxl_prompt_styler]
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?
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 👋🏻
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.
@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?
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.
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.
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?
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?
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.
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
@@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.
I actually find the pricing of this quite offensive. Especially considering its literally an img2img sampler and upscaler looped many times (no doubt running on beefy huge gpus) but the price is outrageous, as its based on open source.
Classic diffusion 'workflow', start with a base image, trying to get a better version of, like the same person, only higher resolution, find out that's impossible, end up with a similar but different enough high-res version that it's useless if consistency of matters at all (person etc)
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?
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 :)
@@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
#### 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
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!
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
it both upscales and 'reimagines' the image, you're correct.
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.
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
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.
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 ❤️❤️❤️
Insulting prices. It’s not even a real upscale, it’s just img2img.
For god sake... it's more than that...
th-cam.com/users/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 $8000+ cards for the server farms
FYI Invoke AI 3.6 has tiled upscale too and mimics magnify AI for free.
It is not free anymore either
@@Greco1114 ??? did they changed it? I have newest version 3.6 and it's free. Also Fooocus has the same upscaler too (you can change slider preset value when activating debug mode in advanced tab)
Hi@@DanielSchweinert Unless you have something different, I check just a while ago (because I read your message actually) and it asked me to register and once I did it took me straight to purchase a subscription.
Absolutely BLOWN AWAY! I had to stop and comment.
I’ve watched closely the AI scene for the last couple of years. I was STOPPED dead in my tracks at the midjourney 6 upscale. Whaa-the-faaa… That is NOT trivial what it did! Unreal. Truly unreal!!!
Thanks again for everything Olivio
Would upscaling with Magnific before uploading to Runway help to reduce Gen-2's morphing problems with image-to-video? I'm sure it would be noticeable with Luma but Luma is way too expensive as it is without the added Magnific cost.
My friend thanks for that workflows! You are an artist with comfy
Thank you Olivio for sharing the 3 comfyUI workflows. It should help me learn it better. Great video!!!
I finally used the landscape upscaler. It was awesome once I was able to use it. My comfyui had issues with 2 missing installs. I had to substitute for other nodes. Once I was able to get it working, I really enjoyed using it. Thanks for making it.
How do you like the Topaz upscaler? Not free but not subscription based.
I tried multiple upscalers, Topaz works best.
Hi Olivio, where do I upload the upscaler models?
The price is ridiculous. The fact that you are giving it the okay is sad.
Bunch of "Web thieves "
Avoid at all costs ☝️
selling out
It seems simple at first but the flow charts at the was a little confusing. Can you go through that in more details at a later date. Thanks!
I have been looking for somthing like that - thank you for saving 40$ per month:)
I'm learning to focus on upscale, I want to ask, sorry, maybe it's a bit difficult for me to understand.
My question is, can this upscale directly enlarge an image without having to add other programs/notes (such as Comfuyi's simple upscale), because what I noticed in the video only changed some of the image details rather than the core of the upscale itself. Maybe there is a setting for how many times we can upgrade, like 2x / 4x in the notes, if so, that would be very helpful, once it works, it's done!
Perfect timing needed this just downloaded a movie.
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?
not work with SDXL models :( how to use this workflows. sorry im newbie
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.
The AI really knows how to improve any photo or video.
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.
He needs to give himself some more medals
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 ?
what r the workflows used in? what is that program?
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! 🙂
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).
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
Thanks for the workflows Olivio😊🎉
Do the images have to go through Magnific's website or can I do it offline on my PC?
I will never pay for any of these paid AI services that are just taking an open source code and charging for it.
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?
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 !
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.
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'
Thank you so much, how to get Promotion code please?
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. 😂
My shows have never looked better, thanks for sharing!
comfy with ultimate sd upscale is so annoying,still waiting for better free solution
tks for the workflows, are you aware that magnific devide the image in smaller chunks and upscale them apart to later ensamble again?, whats is that technique name?
That's what the Ultimate Upscaler does. That's also the name of the node or plugin
@@OlivioSarikas Is there a way to run it independently in colab? without having to install full programs like automatic 1111
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!
I am getting this error at the kSampler....'NoneType' object has no attribute 'shape'. any thoughts?
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?
This is great!
I've been cracking my skull trying to get more detail without modifying the image.
8:55 is better than 2:21, because it's more faithful to the original. For example, it doesn't add those sharp, unnecessary laugh lines that 2:21 does. The laugh lines make her appear older and thinner, imo, and alters her mood slightly.
Can"t we achieve this with some workflow in comfyui ?
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.
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.
Will this work on an entire body that's blurry or only a face?
as usual awesome presentation and right to the point cool workflows
Can we do it with a video in input ?
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.
I get an error saying
Conflicted Nodes:
ColorCorrect [ComfyUI-post-processing-nodes], ColorBlend [stability-ComfyUI-nodes], SDXLPromptStyler [ComfyUI-Eagle-PNGInfo], SDXLPromptStyler [sdxl_prompt_styler]
Hello Olivio, thanks fro sharing. I have the error import failed with the ultimate sd upscale, please help me!
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?
Does magnific do full resolution previews?
is there a free alternative like for webui a1111
The problem is it's too expensive for an image maker. (,not true upscaler)
No this is actually amazing 🔥🔥🔥
Very interesting workflows! I'll definitely check them out. Thanks, Olivio!
Do I need to download these upscale files in order to use this workflow?
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 👋🏻
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.
@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?
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
привет, а для Автоматик1111 есть урок?
These are great workflows. Thank you Olivio!
Very impressive, thanks again on your workflows - got me to be much more into comfyui
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.
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.
Gonna upscale like crazy now.
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?
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.
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.
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.
I recommend watching the video at 1080p or higher.
Watching at 720, it seems to blur the fine details.
I actually find the pricing of this quite offensive. Especially considering its literally an img2img sampler and upscaler looped many times (no doubt running on beefy huge gpus) but the price is outrageous, as its based on open source.
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?
Your lectures are the best. Thank you always!
Classic diffusion 'workflow', start with a base image, trying to get a better version of, like the same person, only higher resolution, find out that's impossible, end up with a similar but different enough high-res version that it's useless if consistency of matters at all (person etc)
Those 1x upscale models seem new (to me). Have you posted a video about those in the past?
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?
Anyone having issue with getting controlnet aux nodes to work?
Hi. where do i get all the models from the workflow?
Your are THE MAN! 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.
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 :)
What program do you do it in?
Just downloaded, the quality blew me away!
I have one question there's any one pay to use it ?
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
After 4 month there are no new better alternatives?
LOL, 2:36, what is that blue bug on her ear?
What are those prices though? Not even worth to try.
Like the demo of Magnific Upscale but started to get lost with the rest of the video, needs a bit more explanation of what your doing for newbies.
This was really helpful.