*Sup*
Not gonna paste some elaborate AI-generated description here full of bullshit about how cutting-edge this model is or try to steal your buzz.
(but if you wanna donate some that's dope, it does help with all the character trainings I've got in the works.)
Update 7/29 - Updated my node pack to fix a couple things, but added way more functionality as well. New LoRA filtering and looping options, automatic trigger detection and routing to a prefix, can support separate automatic outfits for up to three subjects, automatic organized output using values built from dropdown menus,
Update 7/27- NODE PACK -
I consolidated my massive workflow down into 5 brand new nodes. Check them out here, a ton of quality of life improvements and built in Civitai metadata that works easily. Would love to know how functional they are for everyone. Install from the zip uploaded on Civ, or search sick-ollie in the Nodes Manager in ComfyUI, or from GitHub:
https://github.com/sickollie/SickOllie/tree/main
Update 7/22- Krea2 -
Works best for me at high res - 1440x1920 is my go-to.
Sampler - Euler
Scheduler - Beta
CFG - 1
Steps - 9Import my showcase images for a workflow I've been working on. Tried making it easy to figure out, message me if you have any issues with it.
(also the arrows don't line up until you generate one image with it because the preview should pop in at the bottom, but after that they should point to things correctly.)
My goal was simple: Character LoRA compatibility.
Honestly, I'm mostly uploading this because it's annoying seeing so many models posted for stupid high costs. So if you make something cool with it, I'd appreciate you posting your generations and helping get it some exposure so I can compete with the giant models milking people for buzz with new bunk versions every week.
Z-Turbo specific below:
_______________________
It generates NSFW content pretty damn well with the right prompting.
I'm honestly pretty happy with where it's at, and I don't want to slowly destroy the LoRA compatibility just to keep pumping out new versions to make buzz. But we'll see.
ANYWAYS,
Highly recommend trying the two other Text Encoders on CivitAI. I'll add them to the suggested resources. They can make a surprisingly big difference sometimes. Some of my examples use them, but most were generated with the default Qwen 3_4b.
THM_UNC_Z-image_TE
Z_Image_TE_SoLordZ
Description
FAQ
Comments (60)
Thanks for your hard work. It definitely can handle multiple loras at strengths between 0.5-0.8 without losing stability. great job !
Man, that looks awesome! Thanks for your work!
testing it now. So far, i am impressed. Good work!
I like you. I like what you're doing. I like what you stand for. You keep doing you.
I'm going to test this and let you know bluntly. I don't mind paying for people's things because it does cost money to train and all, but it becomes an issue when people constantly make updates with little changes. Then it becomes a milking method in stead of real improvements. I'm glad you used other people's prompts. This way we can show that we are not using some exclusive secret prompts that only the OP knows.
Man this is nice, great work.
I would very much like to train some loras compatible with this, should I train on turbo or base?
And was this trained on turbo or base?
Sorry for the dumb question.. been out of the zimage loop for months now, your model got me hooked again
Not dumb at all! I'm right there with you haha. This is a turbo base, but from my testing it works really well with Loras from either. All of RLY's newer LoRAs are trained on ZiB and have been my go-to for testing.
For LoRAs it seems training with only a triggerword as the whole caption works best, but I rarely actually use the triggerwords and they still work fine.
@sickollie Noted! So you're training on turbo with ai-toolkits de-distillation or something? or whatever the musubi equivalent is?
@sickollie Oh hold up, this is a checkpoint finetune, not a lora merge? man.. I am rusty
kudos for the description.
i'm sold.
will try this thank you
Very good results. Thanks! Great with res_2s_ode or res_2s with beta scheduler. Is a BF16 available? or lora of it?
This seems isn't a merge is it? I suspect not. Have tried many loras and models, and yours certainly has something special.
I just finished uploading a BF16 variant! I’m still getting back up to speed after taking a break since the SDXL days lol, so hopefully that new file hits the mark. Should be able to choose it from the download area.
@sickollie thank you! it's there. downloading now. from fp8 tests,, it certainly plays nicely with sliders (lorahollic eg for bodyhair, weight, camera angle, luisa_pinguin's age slider), and works with loras for character, clothing, brace etc. Can need a little help for some positions, and low value alcaitiff's mysticxxx can work, but of course bring inevitable face changes.
nice model
I have to agree with the majority here: great work. And me too would like to see a BF16 version.
I just finished uploading a BF16 variant! I’m still getting back up to speed after taking a break since the SDXL days lol, so hopefully that new file hits the mark. Should be able to choose it from the download area.
thx, BF16 would be cool—then I could use my converter to play it on my old 8GB AMD clunker :)
i wonder if bf16 and fp16 will make even better quality?
@Melodic_Possible_582589 For AMD Strix Halo, FP8 is not good, I hope he released a BF16
@Melodic_Possible_582589 @clevnumb @ratwatat I just finished uploading a BF16 variant. I’m still getting back up to speed after taking a break since the SDXL days lol, so hopefully that new file hits the mark. Should be able to choose it from the download area.
@sickollie big thx, bro, really !
@sickollie @sickollie thanks. i'm still testing the 8bit version. what can i expect from the bf16 vs the 8fp?
@Melodic_Possible_582589 From my testing it’s mostly the same outputs. I wouldn’t expect a big quality jump from BF16, it’s more of a compatibility option for workflows/trainings/AMD that don’t love the FP8 file. If the FP8 version is already working well for you, you’re probably good there.
@sickollie best turbo model, definitive
One of the best and most realistic models I have ever used. I am impressed by both quality compatibility and also prompt adherence. Well done on creating something which is my view far surpasses and is superior to most other models! I hope you will consider BF16 in the future. Must give credit where credit is due!
Well dang, that is such a nice compliment! thank you! Would love to see some of your outputs if you get the chance to post any!
Also I actually just finished uploading a BF16 variant. I’m still getting back up to speed after taking a break since the SDXL days lol, so hopefully that new file hits the mark.
Tried BF16. Subtly different, similarly good, but on the fence as to whether it is 'better'. I regenerated some fp8 gens - some gave improvements. For others, BF16 was detrimental, as the extra detail resulted in few where nipples were worse - you know - where model has sort of pasted one over another, so the areola changes color part way. Great to have both fp8 and bf16 so can try both - thank you!
The image quality is excellent 👍
best Checkpoint that I have found so far. The realism is stunning. Will keep it as my favourite.
However there is one point that needs improvement. It only creates NSFW images of females ith labia minora visible. No innie-pussies, no cameltoe-pussies. Even if I throw strong loras against this checkpoint, it shows me labia minora (which I really don't wanna see). Any chance to give Innie pussies a chance with the next version? Would be great!!!
++
Works well with LORAs, which it says on the tin. It also works well with controlnet, which other than this model, only CyberRealistic and Moody do well. And this beats both of them, while still being coherent with LORAs. Great for turning rendered, drawn, or semi-real content into realistic.
Dude I love to hear this. I really need to get back into controlnet, havent messed with that since early SDXL days but this might be the push I needed lol
WTF is this MASTERPIECE!?
I've been using this checkpoint for three days. I have to say that what our boy (girl?) here has cooked up is by far one the most lora-friendly models I've tried. It takes loras like bread to butter—producing solid quality outputs without having to excessively lower the lora strength. I don't know what's in this brew, but it's top-notch, grade A certified, Michellin Star-worthy goodness. Try it now, pendejos, and you'll see.
I saw the comments so I tried it out now I'm here making a comment to how ridiculously amazing this checkpoint is. It's insane how good and compatible it is with loras especially with my character loras where I wouldn't get consistent results from time to time with my go to checkpoints but with this checkpoint it just made everything so easy. Now I can save space and delete the other checkpoints I have cause with my specific use case, this checkpoint alone is all I need, plus its faster to gen with it too. To the creator, I hope a rtx 5090 magically gets delivered to you
This model is my new favorite toy.
Amazing, well done.
First comment on this website, I've been testing a lot of checkpoint for ZIT and yours is better than all the "big hyped" I've tested so far !
Hi Author,
what is the goal of your next version? I hope its more realism and pose/prompt adherence.
hey, at the moment I'm pretty happy with where it's at and I am working on releasing a few LoRAs next.
I looked through your images and haven't seen any posted with my model, do you have any to share?
My main concern is keeping it compatible with LoRAs at 1 strength.
Most of the top creators on here arent actually retraining their models for each version, they're just merging in more models/loras uploaded by others which usually makes their model weaker and then charging 10k for it. But I've tested LoRA compatibility with almost every model that gets uploaded on here and I'm still happy with where mine is at for now.
But I would love to see what you've gotten with my model!
@sickollie Hi, thanks for the reply. I made some but haven't had a chance to select them. I'll upload some within a few days. That makes sense why other's usually break or degrade in quality over time, etc. Can you make a penis size (length, width) slider for one lora?
@Melodic_Possible_582589 I have never actually tried making a sliders LoRA myself, but I did see this one recently by a creator that makes quite a lot of sliders, interested to see if it works because I havent had time to test it yet:
https://civitai.red/models/2687793/penis-size-slider-zit
Just WOW. Just found this a few minutes ago, so early in testing here, but wish I had seen this a week ago. Best way to put it is based on extensive testing for months with all available checkpoints I had come to the conclusion that for all its strengths ZIT was fundamentally incapable of this re LoRA mixing and cumulative strengths. Whatever you did here under the hood, PLEASE share with the community. Phenomenal work.
You have done it you mad man! I've tried a bunch of popular ones, but this by far is the best one. Works great on its own and plays well with character Loras. Well done!
best checkpoint by far, not even need a lora like mystic.. crazy job dude congrats
This is the best ZIT checkpoint with character LoRA compatibility I've tried but the LoRA likeness is still a tag diluted than using the official Z-Image Base with 8-step FunLoRA.
Hello,
In your description, you mentioned "Highly recommend trying the two other Text Encoders on CivitAI. I'll add them to the suggested resources."
But in Suggested Resources, there is only one text encoder ( THM_UNC_Z-image_TE ) and three LoRAs.
Did you mean one? Or did CivitAI pull a CivitAI again?
Hey! That's my bad, I actually thought it was capped at 4 resources there and removed it before adding my LoRAs in lol. I just added it back in but also it's here:
civitai.red/models/2662833/zimagetesolordz
I also added them as clickable links straight into the description so they don't ever get pushed out completely!
Honestly wasnt sure how many people were actually reading that haha
可兼容LORA的模型，太棒了！
best ZImage checkpoint out there ❤️
This is definitely the best ZImage checkpoint that exists right now, it consistently just puts out the best results for whatever I throw at it and its really not even close. I'm so unbelievably impressed with it, its a genuine masterpiece. I just want to communicate the only two real issues I'm having in case you ever decide to do a v2 or in case someone else has managed to figure out how to resolve them.
1. The genital anatomy quality on this checkpoint is legitimately maybe the best out there, but I see it really struggle to consistently give good results for vagina penetration from behind. It always tries to generate the top of the vagina no matter what I try.
2. It just refuses to do big, plump asses and really loves to always just spread those cheeks. I've really tried everything and its just really hard to make it cooperate.
man I just typed so much and accidentally clicked the wrong spot and it all disappeared lmao.
First off thanks! I'm undecided on making a v2 at the moment, really focused this one hard on keeping that base Z magic + LoRA compatibility with decent NSFW capability, because nearly all of the big creators keep releasing 10k Early Access models that just butcher character LoRAs, because all they're doing is baking in more and more NSFW LoRAs which makes it impossible to use your own LoRAs on top of those. Plus so many of those basically look like SDXL models now after they've overcooked their models to oblivion lol.
#1 - without seeing your prompts I have a couple theories that may help. It could be a narrative thing it isnt understanding correctly or is signaling to it that it needs to generate that. I may be way off here or giving obvious advice so apologies if I am, but things like "if you dont want to see it, don't say it." Like as soon as you say vagina, it's going to try and make a vagina. Especially if its earlier in the prompt. So you kind of need to say things without actually saying them. And just throwing it out there, ChatGPT can be great for helping with that because ChatGPT cant actually say a lot of those things lol.
Another little thing I noticed recently is something like this prompt section:
"Her right hand rests casually on the BMX handlebar, while her left hand rests on her hip,"
I kept getting her left hand crossing and going to the handlebar. Her left hand was also closer to the viewer in the image so it was more prominent, so flipping it like this:
"One hand rests on her hip, while her other hand rests casually on the BMX handlebar"
fixed it, because it processed the "back" hand second.
So this stuff is full of weird little shit like that lol.
#2 - I'd actually love to see if I can get this version to help with that, so if you have a test prompt to share I can see what I can come up with. That said, I have had similar issues with similar concepts myself. Like weight in general seems to cap out around a certain heavyset level. Heavily prompting for it can sometimes get you what you're trying for, but usually only in a handful of outputs.
Alternative text encoders might help also for both of these issues, I linked a couple in my model description if you want to try those out. I almost never use Qwen 3_4b for anything anymore.
But also I may see if I can train some concept LoRAs for some of these things. I've been trying to figure one out for "innie vaginas" too for another request lol.
Thinking of doing the same for Krea2?
I have been thinking about it, but I still have a lot to release for Z at the moment and may end up updating this model. I'm about to head out of town though so it may be a couple weeks before I could think about starting on Krea2.
Seems like a lot of people have been jumpin to that ship.
it's released! lmk how it does
@sickollie oh wow, nice! ill try it this evening. i forgot this was a whole model. i am one of the crazy people using the whole krea 2 full size model. any chance of this being a lora?
yo it is pretty good
Where's the Vanah character Lora? like cmon... ?
Hey, Vanah is my character. I'm reworking the dataset to get rid of the oily skin issue. I'll be making it available for ZIT and Krea2 in a day or two.
Thank you for this wonderful model. My impression is that, compared to other models, this one excels in skin texture and the representation of ethnicity, though its anatomical accuracy is slightly less refined. I hope this feedback proves useful for the development of v2.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.