This is a mix of many different Models, to play around with the model merge feature of aiNodes which once again has produced a wonderfull result.
The Models in this mix and in that order are
deliberate_v11, babes_11, elldrethsLucidMix_v10, woopwoopPhoto_12, hassanblend1512And_hassanblend1512
They all got mixed at 0.5 one after the other.
There is plenty triggers which can be found in the original models. To show the mix brings together the goods of all the Models I reproduced a few of their Prompts. The results look different for sure but they stilll show the originals features.
aiNodes can be found here: https://github.com/XmYx/ainodes-pyside
Find our Discord: https://discord.gg/VRqwu89Vrw
Description
Added a ckpt
FAQ
Comments (15)
Since this is based on the original Deliberate, please let me know if you run into issues using tools other then aiNodes or auto1111. I will then create a version based on the invoke enabled model.
Can you please add a ckpt version...? Not every setup can use safetensors...
I'm on my way to do so =)
CKPT is now available, dont ask my why it is tagged as pickle tensor. I dont know what that is and how to produce, I can only choose safetensor or ckpt in my tool = )
@osi1880vr That's the way they call it on civitai, don't know why, it is confusing... if you download "pickletensor", you get the ckpt file...
Thanks a lot...!
@osi1880vr the "pk" in ckpt stands for pickletensor. IIRC, CKPT means ChecKpoint PickleTensor.
Some people, including myself found WoopWoop-Photo1.1 to be better, 1.2 can sometimes with some settings overcook images or hard edges and contrast also it seemed to have some odd placement of details issues (I left details on that model) 1.1 had beautiful colors. I grabbed it to try and make better detailed people and prompt stirred it away from NSFW stuff
Have you considered mixing Realistic Vision with it? I have been going back and forth on deliberate and realistic for...well realistic images :) they both have strengths. I like Realistic for simple prompts and deliberate seems to be able to do most tings if you prompt it right. They also seem to make a lot of crossover similarities so should I assume complement each other well, maybe it is overkill though 2 models for the same thing.
I can do, so you suggest a mix of just the both?
@osi1880vr Either that or maybe just adding it to your mix to pull some extra realism. I assume it could bring the best elements of both those models. when I saw you had deliberate it felt (to me anyway) like a natural pairing.
I could do that but it would change what this mix is, so im thinking a new mix would be more to the point. im thinking 3 models woopwoop11 and deliberate and realistic.
@osi1880vr Sounds like a plan!
@osi1880vr I will give that a try. I am just mid system cleanup I tried to add the kohya stuff but decided to try and stick to conda and not venv. Made my env but for some reason all the requirements installed in my main python folder (doh!) so once I have cleaned that mess I will give this a whirl. Looking forward to it! Might mean less model swapping for me. Hoping it will be proto2.2 and this one.
@creeduk So we need to mix in proto2.2 too? :D
@osi1880vr LOL well maybe it is still one of the best goto all rounders. I think still better than 3.4 and 5.8. Have you played with them?
Another interesting one to look at that bestjammer merged into hasdx models was https://civitai.com/models/3762/general-purpose-model which is thought to be a copy of https://civitai.com/models/2781/vintedois-diffusion-v0-1 which makes some great hyperreal images (check user images)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.