Krita Ai, settings for juggernaut v9 lightning model, fast and amazing, even on my GTX1070

Подписаться 3,2 тыс.

Просмотров 5 тыс.

50% 1

Krita Ai, setting up juggernaut v9 lightning model, fast and amazing, even on my GTX1070
Note for correction to video,
Recommend was ( DPM++ SDE ) Via CivitAI , i used ( DPM++ 2M ) in video as no ( DPM++ SDE ) in Krita, I failed to test DPM++ SDE Karras that give fantastic results at low steps at 5...
juggernaut v9 lightning mode
civitai.com/models/133005/jug...
The Krita AI Diffusion maker (Please head over and thank them)
/ @aclysia
Adding another Model directory
• Krita AI, Add Another ...
adding models to krita
• KRITA AI Diffusion FRE...
Krita AI Diffusion installing and over view
• KRITA AI Diffusion FRE...
Olivio, (channel)
removed as ruddy blocked my comments. so will no longer help them.

Хобби

Опубликовано:

26 фев 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 23

@somecallmerob 5 месяцев назад

very cool, playing with the model now using a prompt i was using last night to try to get a nice dark silhouette with clean lines in front of a fire (not alot of luck, occasionally it puts out a good one). I have a 2070 super and it definitely renders images way faster, 5 steps it does in around 5 seconds, and 50 still only takes a little over 10 sec, for a 768 x 1024 image, still not much luck with clean silhouettes hehe, but I'm still stabbing in the dark to get used to it all, thanks for the vid!

@streamtabulous 5 месяцев назад

Yer mastering prompts is very difficult especially because each model might have different words for the trigger. Could try things like shadow person. Or render the fire background image and then do the method showen in second last video of adding a character to a background and prompt that figure for the soild shadow person. It's a great model for sure, but yer different models can certainly work better for some prompts.

@somecallmerob 5 месяцев назад

@@streamtabulous for sure, seems a bit of a balance between keeping it short and adding more information alot of the time. i do plan on splitting backgrounds and characters up for the project i got in mind, but yeah I was just kinda pushing the silhouette thing coz i noticed it was having trouble with that, and just to practise and try out diff settings, the last pile of previews i did were getting way better. had some cool unexpected fire crazyness when i added some faetastic details lora, it also did crazy, unwanted, but awesome things to some space art i was playing around with, i like it, i think the insane randomness is a good way to get ideas too.

@agelessstranger964 5 месяцев назад

A little tip from me on upscaling. I get frustrated with it because it upscales everything(all layers) and seems fiddly to go back down to make possible changes. Now, I just copy the layer I want to upscale into a new document and I don't have to worry about that. I know it isn't much of a tip, but I only just thought to do it.

@streamtabulous 5 месяцев назад

its a good tip, as if you are doing just a in paint section for details on a AI render its clever. for portraits people photos etc i have Topaz photo and gigapixle old version as to broke these days but they do great and faster, they do smooth and blend though so lose details like small skin imperfections. you idea is handy.

@KandooAI 5 месяцев назад

Great Video, awesome to see that it works on an older GPU :) Saw that u used DPM++ 2M which i didnt recommend in my settings ^^ I recommend DPM++ SDE , its a bit slower but had the best quality overall. Euler is also nice and faster, but it lacks a bit in details

@streamtabulous 5 месяцев назад

is that DPM++ Karras (missed trying that) as krita only has the karras, im still working on differences on samplers, Euler is beautiful its definitely softer details but depending on the readers fantastic, I love The the lighting model. will be using it in so editing of photos as the skin imperfections are the best in any model i tested.

@streamtabulous 5 месяцев назад

Edited description to added that error in video, krita has DPM++ SDE Karras tested and fantastic, is the non Karras different or is it just naming in krita that's different for the link to the sampler? as No sampler that's just DPM++ SDE in the Krita diffusion interface. Either Way its working beautifully with other samplers and was good to show the effects of the samplers on a image. I love the result over the over freckles as that sampler is very handy to give say more grit to a zombie for example.

@KandooAI 5 месяцев назад

@@streamtabulous The Non Karras Version is indeed different. Overall the Karras one has the best Quality...but on a few samples i saw some artifacts (not on all). So if you encounter them with the Karras Version then you can switch to the non one

@streamtabulous 5 месяцев назад

@KandooAI yes it's odd that Auspicious Ai diffusion has some samplers not there considering its comfy, must have something to do with the work flow so thr live paint etc works maybe. Not a issue it's clear the models work great even without the identical intended settings. Can't wait to see more models they keep getting better and faster even on this old clunker.

@braintify 5 месяцев назад

Thanks for the content, you're great! It would be wonderful if you would consider the topic of animation in Krita.

@streamtabulous 5 месяцев назад

that is something i need to play with for sure after seeing Acly do lighting

@karmitfrog7968 3 месяца назад

how to connect krita with external sever remotely kaggle notebook . i install stable diffusion on kaggle notebook

@nickolaygr3371 5 месяцев назад

Hi sir mister, can you help with the question? Is there such an option in Krita? For example, I will select 100 by 100 pixels and want the neural network to generate this square with a resolution of 1000-1000 and then reduce it and insert it into a 100 by 100 square.

@streamtabulous 5 месяцев назад

there is no way at current, the selection will be pixels of the selection. one follower mentioned they select the area copy to a new image resize then they do the work and past it back to original layer and resize to fit, too get the higher details. myself i resize the whole image then work on the fine small details at the end.

@jindongzhu8459 3 месяца назад

Hello! May I ask: Why can krita only load XL Model Checkpoint？Thank you！

@streamtabulous 3 месяца назад

Under the configuration where you install stuff it means the sdxl model was unticked. Go through and tick everything the install will re appear and you can install those missing bits.

@jindongzhu8459 3 месяца назад

Thank you!

@beveresmoor 5 месяцев назад

I haven't update this krita plugin for awhile. 😂 I'm still struggle with how to update it.

@streamtabulous 5 месяцев назад

Just takes a bit as needs to download. I back up the old version just incase so the entire 25gig copy the pykrita folder to another drive then do it, just to be safe. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-g3moOJ15jho.htmlsi=fqPRldCQBmgXf679

@beveresmoor 5 месяцев назад

@@streamtabulousThank for your reply. I connect with a custom Comfy UI local server with a regularly update. But I didn't update ai diffusion plugin of Krita. 😅

@streamtabulous 5 месяцев назад

@@beveresmoor yer I have noticed that, only part works, thing like the batch amount etc are missing doing it that way. i been thinking of dropping the run.bat into a back up of the krita installed on just to see if it launches and whats under the hood. as the file structure seems the same just missing some thing to launch it. its why i have not done a video on linking to a comfy install due to the headaches, I'm not smart enough to figure it out yet. you can install the krita one in a different drive through so thats handy for those with no C drive space.