v19 is amazing!!
thank you for your great work sir, model is really good!
Yes, thank you very much for v.19 from me as well 😃
Hi, do you know how to produce the best result while using v19. I just got similar result as v18
Anyone notice that using Step 6-8 using er_sde/beta., it adds a ton of moles on the skin?
Anyone notice that using Step 6-8 using er_sde/beta., it adds a ton of moles on the skin?
Higher AuraFlow shift can reduce most of that, try some thing around 9.5 see if that help.
Anyone notice that using Step 6-8 using er_sde/beta., it adds a ton of moles on the skin?
Higher AuraFlow shift can reduce most of that, try some thing around 9.5 see if that help.
Very interesting. That fixed it.
Higher AuraFlow shift can reduce most of that, try some thing around 9.5 see if that help.
Sorry, i m new to this: what is Auraflow? and if it is a node, how should i link it to the workflow?
thanks
Add a ModelSamplingAuraFlow node like you would with a Load Lora node in between the Load Checkpoint and Ksampler nodes.
Yes, I decided to set it on v19, I can't pick out a problem with this version
v19 is really working perfectly!
This is definitely the best model so far.
As I said in the previous version v18, this v19 crushed the original qwen edit 2511 bf16.
Impressive realism and the checkerboard pattern has been fixed.
Congratulations on the work, which only gets better, just like wine.
prompt used: transform the girl into realistic
no loras, no pos-production, thats a raw output.
Anyone notice that using Step 6-8 using er_sde/beta., it adds a ton of moles on the skin?
Higher AuraFlow shift can reduce most of that, try some thing around 9.5 see if that help.
On the contrary, I think this is an expression of realism
Is the speed of V19 slower than V18?
I wouldn’t say that. But when I run multiple workflow and I have 10 in a raw, average is around 50 sec per image. Various resolution above 1000px. I.e. 1400x1800 . If you do 720 px range it should be around 40sec or less.
For The two versions compared I would say they are very similar.
It depends also on the loras loaded in the workflow and the complexity of the prompt.
ATM I only use F2P Lora , suggested on this forum for face to person adaptation and it works great. I m still testing v18 and 19 and tbh I don’t see much of a difference between the two.
However V18, 18.1 and V19 don’t eliminate each other, they each have certain quality depending on the image you working on.
I will stop yapping now.
Cheers
Anyone notice that using Step 6-8 using er_sde/beta., it adds a ton of moles on the skin?
I've noticed that as well. When using the NSFW version and attempting to add pubic hair, I've seen black, fungal-like artifacts appearing across the character's skin, which is quite unsettling. Additionally, there's a high probability of misplaced hair clusters appearing on the buttocks, which looks very unnatural.
Still using V19, even though at the time of this writing, v22 being out.
Still using V19, even though at the time of this writing, v22 being out.
I am now using back v18.1