Thank you for your information. Quite interesting and unfortunately disgusting what is going with that buzz. Quality will definitely go down if good creators are leaving and mergers are popping up. I've seen multiple new accounts with only 40-50 images ... and the images were almost the same. I was just wondering but now I know the reason for multiplying these "fake" accounts.
I am off to try shakker ai's models in Krita if possible. The model I am looking at says comfyui only, but Krita is using comfy so will it work? I'll report back once the model downloads.
Yer there not just comfy. They probably tested and found better results 9n that and possibly the text image layer. But yes will work in krita and probably fooocus and I'd be shocked if a1111 can't run those models. But yes will work. There a model named normal that's fantastic there also
The idea that I gave them was to trade the art like NFTs or shares in a stock. But unlike meme coins, art can have real value. And those LORAs can really offer value. Just put all the hard work first. All the bells and whistles.
What am I willing to pay for when it comes to art? If there was an easy way to pay for the LORAs I would pay for them. Not a lot, but a buck to $5. I think many would. A good artist should get something. Basic LORA a buck, really good ones 3 to 5. Thank for you honesty and well thought out info. I wish I could talk to you privately.
I guess most people follow a creator to get notifications. Which is how I use it. And to find models I like back easily. The website would be harder to navigate if I couldn't follow creators. Or even those whom create guides / workflows.
for me I'm always looking for newer different trained models to try out, those creators are pushed down already, dice mentioned he paid 50,000 buzz for his model to be featured and then was out bided, so civitAi are featuring models where creator's pay them. that i did not know. not good
I think this is a bad take on SD3. The terms of SD3 license mean that if you make money from TH-cam AD revenue, using SD3 without paying for the license could be a breach. Making money indirectly or directly is covered in the wording. So you yourself already need to apply for a license right this second to avoid potential strikes on the channel, or avoid using SD3 in any way that could gain you monetary returns.
then anyone using it on there video can just turn of monetization off for that video and not have a issue. I think its worse being kept from the public via sites as it stopped people training it and getting to people for personal use. plus its a great learning tool so others can see the training and methods in it and what others do and all that can be taken and learned from for use in training non sdxl3 models. I make no money from youtube so I should probably see if i can get my hands on SDXL3 and test in a video to show people it, since it sounds like there no videos on it but I have seen a number of youtubers using it and It seems fine once there worked out how to prompt with it as it doesn't support traditional prompting that caused a lot of glitches
@@streamtabulous First of all, most people aren't going to make videos using SD3 in case that the rights are purchased by another company and the terms are retroactively modified to impact more wide range of users. Second of all, as one of the most prominent creators on Civitai, their lack of acceptance of it on their platform is NOT stopping people from being able to choose to download, train or use this model. Simply download/upload the model from Hugging Face. Thirdly Shakka is a chinese competitor site, which isn't beholden to the licensing rules due to international differences. And fourthly, the reason that Shakka is such "good competition" in the SD3 space is because they took advantage of Civitai's decisions and their own lack of legal accountability and offered $300,000 to a lucky SD3 model creator who uploads to them.
Also I should mention that in my opinion we need MORE people getting into merging and learning how to improve models. While training is important, merging trained models is the single most important step towards getting higher quality and more prompt adherent models. And most trainers seem to neglect this step, hence why their models are all over the place and require different CFG for different artistic styles etc. instead of having defined spaces for each.
@@tripleheadedmonkey6613 i definitely need to read up on merges more, i did a few in a1111 very easy to get styles and looks i wanted, Wonder if better dedicated programs is out there for merges to show and teach, any recommendations besides a1111
@@streamtabulous Personally I use ComfyUI to merge my own training and custom models, with other peoples lora and custom models. My goal, all round, is to push the envelope by combining models of different styles and niches in small parts in order to build a larger context for the models. By taking 10 models, modifying each with stacks of lora, and then merging them into a single model you can create singular models with the capabilities to excel in all of the constituent models style niches. Almost without drawbacks. In fact in large part if you get the ratio and mixture of styles correct - it will actually fix issues the model has in areas you don't expect.
Model Merges, yes they appear cheap, and yes you can merge two models, but to get a Good Merge? Try at least 3 models, and a few days/weeks testing them, remerging, retesting. imo the only thing harder than getting a Good Merge is training a new model. Chances are that most of the popular models featured in 2023 were merges. Case in point, try to find a version of Juggernaut XL post v6, without Dreamshaper XL merged in, I'll save you the trouble, it doesnt exist. Realistic Vision, Nightvision, and the list goes on, and on.
Not sure any of this matters to me so far. I am commenting as the video plays, but I give no fucks about Buzz. I am only watching to see if I can still download all the models I want. I will check out competitors when we get to that point. I do hope Buzz fails because the way you describe it, it sounds like cosmetics in a game and I want nothing to do with dressing up a digital character or online profile.
a couple of points, yes shakker has sd3, also all the other models they steal and put on their site without permission. they don't advertise it but it's a chinese group, they don't care about licenses or western laws, so there's nothing stopping them. you seem to know how business works, maybe offer civitai a way to monetize, keep good creators around and not hemorrhage money until they die and leave nothing but the chinese sites.
I did not know that. The a user brainslug that posted a link in CivitAi of the same model posted a lot of times. I'd not be shocked if CivitAi AI has models downloaded from elsewhere renamed and uploaded on CivitAi. I don't think we can get away from those practices. Only can hope to find creators that are legit 8n both groups.
@@streamtabulous I have some LORAs posted on another site and Shakker has taken them off that site, without my permission, and placed them. I have contacted them multiple times with takedown requests and they have completely ignored me.
@@10th_Doctor yer since doing this i have learnt there are users taking loras models from civitai and posting them on other sites and vis versa taking claim, i don't believe those running shakker are English speaking/reading so that could be a issue too.
Maximum solidarity, I'm close to you, and I hope you can solve every problem and continue to help us.
hoping you feel better soon enough. Thanks for the insights.
thanks, 34min talk i almost passed out at the end lol cure my rambles, needed to say something though
Thank you for your information. Quite interesting and unfortunately disgusting what is going with that buzz. Quality will definitely go down if good creators are leaving and mergers are popping up. I've seen multiple new accounts with only 40-50 images ... and the images were almost the same. I was just wondering but now I know the reason for multiplying these "fake" accounts.
Thank for your opinion and review of those news. There are so many things happened at the same time for me to keep track. 🤔
I am off to try shakker ai's models in Krita if possible. The model I am looking at says comfyui only, but Krita is using comfy so will it work? I'll report back once the model downloads.
Yer there not just comfy. They probably tested and found better results 9n that and possibly the text image layer. But yes will work in krita and probably fooocus and I'd be shocked if a1111 can't run those models.
But yes will work.
There a model named normal that's fantastic there also
@@streamtabulous I get an error when I try to run it in Krita.
@@agelessstranger964 that's odd I'm not having issues. What's the error?
@@agelessstranger964 I'm now running a number of models from the my favourite is just called normal and is a sdxl model from shakker
The idea that I gave them was to trade the art like NFTs or shares in a stock. But unlike meme coins, art can have real value. And those LORAs can really offer value. Just put all the hard work first. All the bells and whistles.
What am I willing to pay for when it comes to art? If there was an easy way to pay for the LORAs I would pay for them. Not a lot, but a buck to $5. I think many would. A good artist should get something. Basic LORA a buck, really good ones 3 to 5. Thank for you honesty and well thought out info. I wish I could talk to you privately.
I guess most people follow a creator to get notifications. Which is how I use it. And to find models I like back easily. The website would be harder to navigate if I couldn't follow creators. Or even those whom create guides / workflows.
for me I'm always looking for newer different trained models to try out, those creators are pushed down already, dice mentioned he paid 50,000 buzz for his model to be featured and then was out bided, so civitAi are featuring models where creator's pay them. that i did not know. not good
I think this is a bad take on SD3. The terms of SD3 license mean that if you make money from TH-cam AD revenue, using SD3 without paying for the license could be a breach.
Making money indirectly or directly is covered in the wording. So you yourself already need to apply for a license right this second to avoid potential strikes on the channel, or avoid using SD3 in any way that could gain you monetary returns.
then anyone using it on there video can just turn of monetization off for that video and not have a issue. I think its worse being kept from the public via sites as it stopped people training it and getting to people for personal use. plus its a great learning tool so others can see the training and methods in it and what others do and all that can be taken and learned from for use in training non sdxl3 models.
I make no money from youtube so I should probably see if i can get my hands on SDXL3 and test in a video to show people it, since it sounds like there no videos on it but I have seen a number of youtubers using it and It seems fine once there worked out how to prompt with it as it doesn't support traditional prompting that caused a lot of glitches
@@streamtabulous First of all, most people aren't going to make videos using SD3 in case that the rights are purchased by another company and the terms are retroactively modified to impact more wide range of users.
Second of all, as one of the most prominent creators on Civitai, their lack of acceptance of it on their platform is NOT stopping people from being able to choose to download, train or use this model.
Simply download/upload the model from Hugging Face.
Thirdly Shakka is a chinese competitor site, which isn't beholden to the licensing rules due to international differences.
And fourthly, the reason that Shakka is such "good competition" in the SD3 space is because they took advantage of Civitai's decisions and their own lack of legal accountability and offered $300,000 to a lucky SD3 model creator who uploads to them.
Also I should mention that in my opinion we need MORE people getting into merging and learning how to improve models.
While training is important, merging trained models is the single most important step towards getting higher quality and more prompt adherent models.
And most trainers seem to neglect this step, hence why their models are all over the place and require different CFG for different artistic styles etc. instead of having defined spaces for each.
@@tripleheadedmonkey6613 i definitely need to read up on merges more, i did a few in a1111 very easy to get styles and looks i wanted, Wonder if better dedicated programs is out there for merges to show and teach, any recommendations besides a1111
@@streamtabulous Personally I use ComfyUI to merge my own training and custom models, with other peoples lora and custom models.
My goal, all round, is to push the envelope by combining models of different styles and niches in small parts in order to build a larger context for the models.
By taking 10 models, modifying each with stacks of lora, and then merging them into a single model you can create singular models with the capabilities to excel in all of the constituent models style niches. Almost without drawbacks. In fact in large part if you get the ratio and mixture of styles correct - it will actually fix issues the model has in areas you don't expect.
Model Merges, yes they appear cheap, and yes you can merge two models, but to get a Good Merge? Try at least 3 models, and a few days/weeks testing them, remerging, retesting. imo the only thing harder than getting a Good Merge is training a new model. Chances are that most of the popular models featured in 2023 were merges. Case in point, try to find a version of Juggernaut XL post v6, without Dreamshaper XL merged in, I'll save you the trouble, it doesnt exist. Realistic Vision, Nightvision, and the list goes on, and on.
definitely great merges with testing and tuning, but i feel civitai move will lead to one click upload and done.
Are you going to make a FLUX ai tutorial?
i lost two family members one this week. was going to but can't composes myself at this time.
will when up to it but probably a month I'm not sure
@@streamtabulous sorry to hear the bad news. Condolences.
Not sure any of this matters to me so far. I am commenting as the video plays, but I give no fucks about Buzz. I am only watching to see if I can still download all the models I want. I will check out competitors when we get to that point.
I do hope Buzz fails because the way you describe it, it sounds like cosmetics in a game and I want nothing to do with dressing up a digital character or online profile.
I agree, it's very odd strange and messed up
a couple of points, yes shakker has sd3, also all the other models they steal and put on their site without permission. they don't advertise it but it's a chinese group, they don't care about licenses or western laws, so there's nothing stopping them. you seem to know how business works, maybe offer civitai a way to monetize, keep good creators around and not hemorrhage money until they die and leave nothing but the chinese sites.
I did not know that. The a user brainslug that posted a link in CivitAi of the same model posted a lot of times. I'd not be shocked if CivitAi AI has models downloaded from elsewhere renamed and uploaded on CivitAi.
I don't think we can get away from those practices. Only can hope to find creators that are legit 8n both groups.
@@streamtabulous I have some LORAs posted on another site and Shakker has taken them off that site, without my permission, and placed them. I have contacted them multiple times with takedown requests and they have completely ignored me.
@@10th_Doctor yer since doing this i have learnt there are users taking loras models from civitai and posting them on other sites and vis versa taking claim, i don't believe those running shakker are English speaking/reading so that could be a issue too.