Magnific AI Upscaler Free Alternatives! Krea and Comfy UI Workflows
Vložit
- čas přidán 20. 05. 2024
- ✨ One Click Comfy UI Installer: / one-click-for-ui-97567214
🎨 Generative AI Art Playground: www.pixeldojo.ai/?...
✨ Support my work on Patreon: / allyourtech
💻My Stable Diffusion PC: kit.co/AllYourTech/stable-dif...
Magnific produces some mind blowing results with a level of detail you can't find anywhere else. How does it actually work though? Are there any free alternatives you can use? Check the video for these answers and more!
👾Magnific: magnific.ai/
👾Krea: www.krea.ai/
👾Photo Portrait Upscaler: openart.ai/workflows/18UZFMnp...
👾Portrait re-imagined upscaler: openart.ai/workflows/5l3RcKm6...
👾Comfy UI: github.com/comfyanonymous/Com...
⚔️ Join the Discord server: / discord
🧠 AllYourTech 3D Printing: / @allyourtech3dp
👾 Follow Me on X: / blovereviews - Věda a technologie
🎨 Generative AI Art Playground: www.pixeldojo.ai/?
✨ Support my work on Patreon: www.patreon.com/allyourtech
💻My Stable Diffusion PC: kit.co/AllYourTech/stable-diffusion-build
Thanks for making it clear that what people are calling upscaling really isnt upscaling. Although one could make the argument that real upscaling isnt really real either because its uaing math to estimate where one pixel goes and another doesn't which could potentially change an images narrative. Thats why NCIS shows that "enhanced" photos always drove me nuts. Because that information wasn't in the original image. Makes you wonder what we should alow to be admissible in court i the future because someone could say they used ai to upscale and image and use it against a defendant when what the image represents is fake... Sorry for the rabbit hole comment lol.
These discussions are important. You bring up a great point. If someone uses an AI “upscaler” in a legal setting, that could open up a whole can of worms …
Not free
thanks for this fantastic video. an installation guide for using ComfyUI on a Mac would be sublime!
I’ll bust out the Mac and make it happen!
Awesome!!! Thanks!!! @@allyourtechai
I second that!
Just discovered your channel, subscribed! Great video… I’ve heard about ComfyUI, would love to know more about that (IF it can be done on Macs).
I’ll put together more tutorials (it does run on any M series Mac)
It says the Photo Portrait Upscaler app is offline. Is this a permanent thing or temporary? I just used the run command. Or should I use the cloud function? ( I dont have any comfyui workflows created yet) I Just want to test out the enhancing feature on a few images to see the results. ( I figured run command is the best way to do this-am i correct) ?
I have a guide on installing and using comfyui locally that you can use. I have no idea why it is offline or how long it might be offline.
Thank you for sharing the information :)
bonjour je n'ai pas le filtre photo réaliste
how to get the most similar result in ComfyUI like Krea. if it possible?
I’m working on it :)
@@allyourtechai are you sure that Krea uses SD algorithm?
@@andrewq7125 more than likely. I'm guessing it is a custom workflow from something like comfyui
I really don't know how you do with Krea-AI as I never succeeded to obtain a good definition of my low res photos
Most of the upscalers need some tweaking to get a good result, and they are typically better at some types of images than others.
0:39 Who is this Laura Croft? 😫
That's because they're not "upscalers". They're "enhancers". There's a difference. Krea, for instance, has a slider that lets you reduce or remove the new data inclusion so their enhancer operates as an upscaler.
Correct, a point I made in the video
Thanks a lot for this heads up! Subscribed. I'm new to the AI World, trying to get acquainted with the different tools. I've fumbled onto one of those using Workflows.. In short, they'd be what was called a "Macro" most of the time? In other contexts, batch files of commands... Note - just digged in: Krea is now charging like others, and it's not cheap if you're not just casually using this something like twice a month, and it's not UHD 2K (only FHD I found better plans on some other plaforms). About Stable Diffusion, I've tried a very simple prompt: "a frog having the head of a donkey". It didn't work, their LLM isn't good enough. When that's the case on platforms that charge a high fee, there's a lot of waste of compute time/credits/tokens/whatnot => money.
1:55 ... Typical Maccas Burger ADs !!! Hahahaha
😂
Caught your rif there at the end. Wondering if you saw that early internet meme, All Your Base Are Belong to US. You could replace "base" with "tech" and have a sweet retro alternative. 😆
Haha you are the first person to catch that. I used to have a sub heading (are belong to us) under my logo. Might be time to bring that back.
tech tech tech tech- all your tech tech tech tech - all your tech are belong to us (AI) LOL @@allyourtechai
comfy ui not working with amd 8gb it says giving error gpu limit
Which amd gpu do you have?
What if I have AMD RX 5700XT 8GB?@@allyourtechai
Just seeing all these lines in Comfy makes it terrifying to think about.
It doesn’t get any better when you use it 😂. Big learning curve, but also very powerful
Easier than it looks. Start with the basic workflow. It'll all make sense slowly
kreta is hit an miss with fantasy art... some images gives me more then i asked for and for some it alters the original image waaaay to much.
That’s great info to share. I imagine they would need to change the diffusion model they are using to account for different art styles
it does people and scenery really well, but anything abstract, its going to turn it into a monstrosity. i had an image with a large rusty nail in the foreground and it turned it into some sort of sword sheath, or into twigs lmao. i hope the models improve and it can recognize more things. i dont understand how 'rusty nail' is not a thing that is recognizable, seems like a normal object to me
I would think rusty nail would be pretty easy to pick up as well. For some of these, training a LoRA to handle different subject types could probably increase the accuracy, but that doesn't scale well. If you only do architecture photos though I suppose you could do well training an architecture model for example though.
You don't get a commercial license on kria ai free version.
Great call out!
Hey, could you open more tabs in your browser, please? I'm still able to identify some of the websites you have open. 😂😂😂
😂 it gets better! That’s one of my 6 workspaces, all with a similar number of tabs open.
@@allyourtechai lol. Get yourself under control, mate!
how is it free when the website says $39 a month?
It was free at the time (pretty sure I even mentioned that it was likely not going to last). You can use pixeldojo.ai with the same functionality for $5 month.
@@allyourtechaiIts $10 now
life saver
I think people are getting caught up in this thinking it's something that don't exists. you can already take existing images in an SDXL image to image and just getting the settings right to do this exact same thing. this isn't upscaling it's just regenerating the image with img to img stuff. I'm not saying it's not cool, it's just like al the complicated img 2 img stuff done for you and simplified. I think main problem is them calling it upscaling, it's not just upscaling but you know, companies can lie and say anything these days, the government doesn't go after companies anymore for lying or misleading people.
Yeah, you can do some crazy stuff with various workflows, but that's my point as well. It isn't upscaling. It's somewhat dangerous for people to think it is in fact. Imagine someone trying to use it in a legal setting?
last was bad so bad...
Krea free version is for personal use only from my research. I"m looking for tools to bootstrap my channel and my business. I don't want to pay for services starting out.
You will likely need to run models locally on your own hardware in that case
@@allyourtechai yeah I’m looking into that.
@@allyourtechai Playground AI is free for 500 images a day, lots of models to choose from, image to image as well as text to image, inpainting and outpainting, object removal, upscaling. It's pretty great for free. the catch is you are low on the processing que so you just need to work at night.
THESE ARE NOT FREE BRO, Why your headline wrong?
They were free at the time
Magnific AI is NOT free!
Correct, and at no point did I suggest it was
well obviously..
You wrote "a young adult man" but you aren't actually young. Could have gotten better result just writing "an adult man"
That cuts deep! I'm young in my mind :)
Its not upscale its just generator for new images, usless
Krea has devolved and the new version of the enhancer is rubbish that puts eyes and nipples everywhere it shouldn't or makes blurry images. I thought we left DeepDream in 2015. Their support is non-existent and they don't seem to want to do anything about it. At this point it's starting to feel like a scam and I'm cancelling my sub.
Well, check out my project at pixeldojo.ai. It’s evolving rapidly inexpensive
Title Is Real Tricky So You Have To Read It More Carefully And Author Of The Title Knew That Too.
Magnific is not free dude
I neer said it was. It's the most expensive option i'm aware of.
"This is fantastic". No is not. You are in 2024 is an old technology. 🤷🏿♀️
It is not even that impressive if you aren't a beginer.
Your videos are very good to introduce new tech to people that don't know anything about these technologies but don't sell it like it is something exceptional, because it is not. Adapt your language to your audience but keep facts as facts.
Keep doing you great work.
useless