This model changes the level of detail
Detail tweakers aren't new, but here we go anyway. This model increases the number of intricate lines and the complexity of shading, essentially it is a realism LoRA, the higher the weight it is at, the more features will be present in the image with marginal impact on the composition.
Inspiration
I just noticed there wasn't a detail tweaker for Pony Diffusion XL yet and I decided it was prime time to train one. it was quite easy to train and only took about an hour because this is the first and only version I made, if there is an abundance of issues that I haven't recognized thus far, I will make updates, but as it stands it seems pretty good I think.
Donations
Speaking of training, training models is expensive and I run training on my own private server, if you like what I do, consider supporting the development here!
Key Benefits
allows the alteration of details on a finely controlled level
far more predictable than just prompting for "detailed" and praying to a deity of your choice
allows for more extreme detail than would be possible with mere prompt weighting
Quirks
I would recommend putting "realistic" and "3d" in the negatives for V1 because V1 has a tendency to make some uncanny valley things at higher weights.
V2 seems to function exactly how I would want it to. if you encounter any strange behavior please let me know.
Description
Trained with the new and improved disentanglement method
much better at preserving the content of the original generation
FAQ
Comments (16)
Can you create a version for the SeaArt Furry XL 1.0 model?
I'll look into it. I've been attempting to fix some bugs in the code I am using for training so it might take me a while to get around to it, but I'll put it on my to-do list.
@Yolup Alright, thanks.
@raxit34 I'm working on it now because I got bored of my school projects. I should have a good enough model to upload in like a day or two. I'm hoping I'll have one up tomorrow.
Ok, it is uploaded now, it's not as good as I wanted but it works for now, I'll be training more maybe. their legal disclaimer scares me I guess although it just looks like a basic SD license so idfky they needed to include it, I'm not a lawyer so whatever, I did my best to maintain compliance with the license but if it needs anything else I think I'll delete it and never make another model for it again. it's a neat model but if they are going to make hoops for me to jump through then I'll let someone with a legal team and 1/10th the ability come along and do it instead. (I'm really tired so my superiority complex is showing, sorry if it comes off too aggressive)
Great job! Well done <3
V1 works better for me
What base model are you using for testing and what is the desired effect that you are wanting out of the model? perhaps I could incorporate it into the next one I do. I think V2 is quite good already though, if you aren't using the basic Pony Diffusion V6 XL then I can't guarantee compatibility unfortunately, and if you want something more like a realism slider then V1 will definitely be what you want to go for. V2 specifically attempts to avoid being a realism slider.
will you make a version to support ebara pony 1 checkpoint?
I am unfamiliar with that model and my rudimentary search yielded nothing. I need more information to decide if I will train a version for that checkpoint.
@Yolup I dont know why this model cannot be searched in civitai, but I just saw someone used it and I google searched it myself. It seems to be the best for generating 1990s style anime pictures with that Pony model
e.g. https://civitai.com/images/10520815
the model link : https://huggingface.co/tsukihara/xl_model/blob/a404bd664a06603403592a1cd6964100e9974bd0/ebara_pony_1.safetensors
Negative values still alter the image. The differences in detail you get from this "detailer" are minimal and not at all as presented in the example pictures.
How are you using it and how is it behaving, specifically? I can't fix any issues with it if you don't share recreation steps, it seems to work for me and everyone else who has tried it. I look forward to seeing your feedback.
@LindezaBlue then please tell me how to recreate the issue, as explained in my other response, many other people seem to be using it without issue, I can't fix an edge case if you don't tell me how to recreate it.
@LindezaBlue
I don't observe what you are describing when I use the model.
what base model are you using, what prompt, and what settings, there are so many variables you could have that affect the generation. the base model is huge too. it appears to work on my end. until you can show me how to make it not work, I will assume it does work, as it appears to for thousands of other people.
it's more likely that you are doing something wrong, or unusual than that literally thousands of people use it regularly without issue when it "does nothing" so I again ask, how are you using it, what is your prompt, what are your settings, what is the base model, what effect do you expect from it, etc.
I look forward to hearing feedback so I can develop a better model that meets your needs in the future, but until then I literally can't help you.
Thank you, adds a little more life to the pictures!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.