Hard to be not biased ( I am the Creator of Juggernaut) ;) . RealVisXL is a really good one, cant wait to see where his journey goes :) I prob gonna release Version 4 sometime next week :)
My bad folks, I said that Juggernaut was updated Sep 5, I meant Sep 9th! hahaha! Which one do you prefer between Juggernaut XL or RealVisXL? I posted some pics that has a bit more photorealism in the community tab, check it out here www.youtube.com/@MonzonMedia/community
From what I've seen before although I haven't tried RealVis I think I prefer Juggie, however since you showed Real I'll definitely check it out! Video Suggestion: how about a comparison between really good non XL model and the best XL model you can think of to show render time and quality comparisons? I think that'd be one cool video!
I think most people would but definitely check out RealVisXL. I'm still experimenting with it as well. As for the video suggestion...I'll consider it although I think most people know at the moment SD1.5 is still better for most situations for various reasons....but explaining why could make for some good content too. Appreciate the suggestion bud!
😎 awesome, as always....thank you :D Can you paste the prompt for that tiger and beer? I want to try those with other models and , say invokeai and Fooocus ✅🙂💖
Sure thing boss! Added them to the description so others can try it out too! Pretty simple prompt, you can definitely get a lot more with more details. Have fun!
what are the hardware requirements for Juggernaut? I'm trying to get it run with Stable diffusion but it takes forever to load and often crashes. I'm running linux Ubuntu, RAM 16GB DDR4, AMD RX 6600 8 GB VRAM GFX card, Ryzen 5 3600 processor. Enabled ROCm but still no results. Just wanted to ask what's your hardware?
Hello Sir, I'm new here. I just have a question I hope you can answer. What do you think are the advantages of having stable diffusion running locally on GPU vs just running it on website like Playground Ai? I'm wondering if it's worth it, since Playground Ai already looks really good.
Great question...there are a few things to consider. Running it locally you have full access to all the open source resources, models, controlnet, extensions etc. However it requires a decent GPU to run it, minimum 8GB of VRAM although some local platforms you can run it on as little as 3-4GB however it will be slow. The way things are going 12GB is becoming the minimum. With that being said the costs can add up as GPU's alone are very expensive. Let's assume someone new is looking to run SD locally they would need a computer that meet's the requirements with a decent GPU and enough storage. Costs can vary but let's say a decent set up would cost $1500 us. Whereas using a service like Playground you can use it for free but it's limited in terms or resources and tools. If you buy pro you have full access at $15/mth or $180/yr. No need to have an expensive computer as it's an online service and even after 5 years of subscribing you would still spend less than buying a computer. But the main benefit with running locally is you have full access to all the open source tools available as long as your system can run it. So everyone would value it differently. Hope that helps and let me know if you have any other questions.
@@MonzonMediaThank you very much sir! I'm glad to finally have a good answer to this little dilemma. I've bought RTX 3060 12gb (second-hand) for stable diffusion, but unfortunately I don't have a good PC yet. I'll be trying Nightcafe and Playground Ai for now while eyeing for a good second-hand PC (Intel 12th gen). Still, installing stable diffusion seems a bit intimidating for a newbie like me. There's a lot of learning to be done. I'll be using Ai to generate images for my novel. I'll go slow for now. Again, thank you very much for your well-thought answer! ✨
You're welcome, glad I could help. I'm curious why you got the GPU before getting the PC? Or do you already have an older one and it's not compatible with the GPU?
@@MonzonMediaThat's correct. I have an old PC which keeps running like a champ. Sadly the pcie was actually pcie 2 and not pcie 3 as I expected. 😢 Btw, just returned from buying a second hand i5 12400. I should have asked for cheaper but I was too shy. And now I'm broke. Oh well... 😅 Not gonna cry! Time to set up the new pc...
How can I download this options? I know there is some png which can automaticly build it for me, but it is difficult to find proper. Or maybe I could see the whole model map? Sorry for bad English
@@MonzonMedia yes, workflows XD. Actually, I already found ComfyUI ULTIMATE Workflow 3.2 but got this errors: When loading the graph, the following node types were not found: Evaluate Strings Evaluate Floats Load Face Enhance Model (mtb) Restore Face (mtb) Evaluate Integers Nodes that have failed to load will show as red on the graph. Have you had similar errors? I checked all missed nodes, all was installed.
Not sure my friend, I always try to build my own workflows so I know how to fix it if there is anything wrong. Other than the nodes that need to be installed, likely there are some nodes not connected or even out of date. I have this one that is very simple to start off with if you want to try it. drive.google.com/file/d/1CiOdkh088KspfQcQ2S7UgaTPuzwEzVlD/view?usp=sharing Just download it and drag it on the comfyui workspace.
Will definitely add it to my next SDXL comparison. I can tell you it does really well with the right prompting. I use the base model quite a bit most of the time. 👍🏼
@@MonzonMedia I love that SDXL performs well without trying to force 20 year old sexualized women into every image I produce like all the top tier pre-SDXL models.
Yeah I totally hear that, I think part of it is because SDXL has some SD2.1 data and the data set was a lot more cleaner. Honestly I had to purposely prompt a certain way to prevent it for my examples. I mean people can do what they want but sometimes (especially SD1.5 models) you don't have to try to get those kind of images. 😆
@@MonzonMedia what are the hardware requirements for Juggernaut? I'm trying to get it run with Stable diffusion but it takes forever to load and often crashes. I'm running linux Ubuntu, RAM 16GB DDR4, AMD RX 6600 8 GB VRAM GFX card, Ryzen 5 3600 processor. Enabled ROCm but still no results. Just wanted to ask what's your hardware?
@@Rich_MrI regret that whatever I learned about Linux I have forgotten. But if you are running Stable Diffusion with the AUTOMATIC1111 UI then 8 GB VRAM is marginal. 12 GB VRAM is really the working minimum. Also, you are using an AMD series video card. Stable Diffusion supports AMD but it has been tuned to NVidia, so that is a disadvantage. At 6.9 GB, Juggernaut XL is a little larger than many community models (including RealVisXL), which are about 6.7 GB. Possibly that is just enough to put you into the crash zone. But if you use the Fooocus UI then you can use as little as 4 GB of VRAM, which is what I have. You should have no problem running Juggernaut, the default model for Fooocus. It is slow for me, but very stable. I am using 28 GB of system RAM, so that does help compensate for the shortage of VRAM.
Hard to be not biased ( I am the Creator of Juggernaut) ;) . RealVisXL is a really good one, cant wait to see where his journey goes :)
I prob gonna release Version 4 sometime next week :)
I'm honored sir! Appreciate you visiting and I look forward to seeing V4!
My bad folks, I said that Juggernaut was updated Sep 5, I meant Sep 9th! hahaha! Which one do you prefer between Juggernaut XL or RealVisXL? I posted some pics that has a bit more photorealism in the community tab, check it out here www.youtube.com/@MonzonMedia/community
From what I've seen before although I haven't tried RealVis I think I prefer Juggie, however since you showed Real I'll definitely check it out! Video Suggestion: how about a comparison between really good non XL model and the best XL model you can think of to show render time and quality comparisons? I think that'd be one cool video!
I think most people would but definitely check out RealVisXL. I'm still experimenting with it as well. As for the video suggestion...I'll consider it although I think most people know at the moment SD1.5 is still better for most situations for various reasons....but explaining why could make for some good content too. Appreciate the suggestion bud!
@@MonzonMedia you're welcome and hey no worries on suggestions they're just that you do you, you're doing great and we'll keep enjoying!
This was great Make more comparison videos
thank you
I have a few coming soon! Stay tuned!
😎 awesome, as always....thank you :D
Can you paste the prompt for that tiger and beer? I want to try those with other models and , say invokeai and Fooocus ✅🙂💖
Sure thing boss! Added them to the description so others can try it out too! Pretty simple prompt, you can definitely get a lot more with more details. Have fun!
what are the hardware requirements for Juggernaut? I'm trying to get it run with Stable diffusion but it takes forever to load and often crashes. I'm running linux Ubuntu, RAM 16GB DDR4, AMD RX 6600 8 GB VRAM GFX card, Ryzen 5 3600 processor. Enabled ROCm but still no results. Just wanted to ask what's your hardware?
Hello Sir, I'm new here. I just have a question I hope you can answer.
What do you think are the advantages of having stable diffusion running locally on GPU vs just running it on website like Playground Ai?
I'm wondering if it's worth it, since Playground Ai already looks really good.
Great question...there are a few things to consider. Running it locally you have full access to all the open source resources, models, controlnet, extensions etc. However it requires a decent GPU to run it, minimum 8GB of VRAM although some local platforms you can run it on as little as 3-4GB however it will be slow. The way things are going 12GB is becoming the minimum. With that being said the costs can add up as GPU's alone are very expensive. Let's assume someone new is looking to run SD locally they would need a computer that meet's the requirements with a decent GPU and enough storage. Costs can vary but let's say a decent set up would cost $1500 us. Whereas using a service like Playground you can use it for free but it's limited in terms or resources and tools. If you buy pro you have full access at $15/mth or $180/yr. No need to have an expensive computer as it's an online service and even after 5 years of subscribing you would still spend less than buying a computer. But the main benefit with running locally is you have full access to all the open source tools available as long as your system can run it. So everyone would value it differently. Hope that helps and let me know if you have any other questions.
@@MonzonMediaThank you very much sir!
I'm glad to finally have a good answer to this little dilemma. I've bought RTX 3060 12gb (second-hand) for stable diffusion, but unfortunately I don't have a good PC yet.
I'll be trying Nightcafe and Playground Ai for now while eyeing for a good second-hand PC (Intel 12th gen). Still, installing stable diffusion seems a bit intimidating for a newbie like me. There's a lot of learning to be done. I'll be using Ai to generate images for my novel. I'll go slow for now.
Again, thank you very much for your well-thought answer! ✨
You're welcome, glad I could help. I'm curious why you got the GPU before getting the PC? Or do you already have an older one and it's not compatible with the GPU?
@@MonzonMediaThat's correct. I have an old PC which keeps running like a champ. Sadly the pcie was actually pcie 2 and not pcie 3 as I expected. 😢
Btw, just returned from buying a second hand i5 12400. I should have asked for cheaper but I was too shy. And now I'm broke. Oh well... 😅 Not gonna cry! Time to set up the new pc...
@@FielValeryRTS That card can run in a PCI-E 2.0 slot (I'm doing exactly that myself), so, just curious why you said that it wasn't compatible?
I prefer Realvis this round, Juggernaut can often be blurry due to the merges?
Both are very good models and it really comes down to personal preference. But you may have a point on the merges.
How can I download this options? I know there is some png which can automaticly build it for me, but it is difficult to find proper. Or maybe I could see the whole model map? Sorry for bad English
I mean some templates, maybe?
The workflow for Comfyui?
@@MonzonMedia yes, workflows XD. Actually, I already found ComfyUI ULTIMATE Workflow 3.2 but got this errors:
When loading the graph, the following node types were not found:
Evaluate Strings
Evaluate Floats
Load Face Enhance Model (mtb)
Restore Face (mtb)
Evaluate Integers
Nodes that have failed to load will show as red on the graph.
Have you had similar errors? I checked all missed nodes, all was installed.
Not sure my friend, I always try to build my own workflows so I know how to fix it if there is anything wrong. Other than the nodes that need to be installed, likely there are some nodes not connected or even out of date.
I have this one that is very simple to start off with if you want to try it. drive.google.com/file/d/1CiOdkh088KspfQcQ2S7UgaTPuzwEzVlD/view?usp=sharing Just download it and drag it on the comfyui workspace.
@@MonzonMedia Thank you!
I assume these will work with Automatic 1111
Yes any stable diffusion platform. 👍
I'm really curious how base sdxl compares to these...
Will definitely add it to my next SDXL comparison. I can tell you it does really well with the right prompting. I use the base model quite a bit most of the time. 👍🏼
@@MonzonMedia I love that SDXL performs well without trying to force 20 year old sexualized women into every image I produce like all the top tier pre-SDXL models.
Yeah I totally hear that, I think part of it is because SDXL has some SD2.1 data and the data set was a lot more cleaner. Honestly I had to purposely prompt a certain way to prevent it for my examples. I mean people can do what they want but sometimes (especially SD1.5 models) you don't have to try to get those kind of images. 😆
Plz tell how to run this on wondows 10 pro and it is free?
What are your system specs, especially your GPU?
@@MonzonMedia what are the hardware requirements for Juggernaut? I'm trying to get it run with Stable diffusion but it takes forever to load and often crashes. I'm running linux Ubuntu, RAM 16GB DDR4, AMD RX 6600 8 GB VRAM GFX card, Ryzen 5 3600 processor. Enabled ROCm but still no results. Just wanted to ask what's your hardware?
@@Rich_MrI regret that whatever I learned about Linux I have forgotten. But if you are running Stable Diffusion with the AUTOMATIC1111 UI then 8 GB VRAM is marginal. 12 GB VRAM is really the working minimum.
Also, you are using an AMD series video card. Stable Diffusion supports AMD but it has been tuned to NVidia, so that is a disadvantage.
At 6.9 GB, Juggernaut XL is a little larger than many community models (including RealVisXL), which are about 6.7 GB. Possibly that is just enough to put you into the crash zone.
But if you use the Fooocus UI then you can use as little as 4 GB of VRAM, which is what I have. You should have no problem running Juggernaut, the default model for Fooocus. It is slow for me, but very stable. I am using 28 GB of system RAM, so that does help compensate for the shortage of VRAM.