@Stable_Yogi I have pretty much the same problems. Even if i copy the seed, sampler and prompts etc. over, the images I get don't look anything like the original. I have been using plenty of sdxl models without problems.

@Stable_Yogi Actually, I am also using A1111. I will try a few different things. Running it through all samplers and schedulers.

@Stable_Yogi Forget what I said. I am stupid. I downloaded the wrong model. I used the pony version instead of the SDXL version. The SDXL version actually produces good images.

@Stable_Yogi 
I do, but about all I can see differing is minor usage info like suggested samplers, steps, and CFG. Then even that info is something of a pain to get as it is in no way highlighted, so we need to have a tab open for each subversion, and carefully read each part, by part, comparing them.

Although even then, even if you see those things, there's no indication of WHY those differences exist. For instance, you used to indicate the DMD2 ones, were based off of DMD2. Well great, but there was no indication as to what DMD2 even was, and lately you're not even stating that much. After doing a bit of a search on DMD2, I guess you mean it was based on a more porn related model, but since you made no clear indication, I can only assume this is the case, which itself is part of my point. Also you just recently made an ALT version of the DMD2 subversion, but there is no about entry for it, so what is different between the two? Then what's FP16, what's BF16, and why should I care?

You make good models, LoRA, and such, but if you're going to make sub variations of them, please at least clearly indicate what makes one, different from the others, so we can pick which would be right for us, without having to get them all, and test them extensively. I'm sure you know what makes them different, but we can't read your mind, you need to tell us these things directly.

@AImodels4me2use for a little info: BF16 is for data center GPUS, FP16 is for gaming cards.

@Stable_Yogi I agree with @AImodels4me2use, specifically, whats the difference between the last DMD vs DMD_alt ? there is nothing on "about this", I just can see on metadata that DMD is low steps low cfg and DMD_alt is higher cfg and low steps. Is there anything else that can help to figure out what results should we expect by using the different alternatives? thanks, you are doing a superb job but more info is appreciated :)

@Stable_Yogi  As an alternate view, I see no problem with your current naming convention, clear enough I would say. Labeling FP16/BF16 builds separately is concise and for me helpful,

@Stable_Yogi yup it is fine to post as versions as you wish if you think this can be useful to someone. We just need the info. FP16 and BF16 is somehow easy to find googling what's the difference, but some creators just make "alt" or "_b" versions without any explanation and that hurts heheh cheers!

@Stable_Yogi Thanks! BTW the DMD2 Alt fine tune CFG, is a really good step up and very interesting

@rollerinot 
Doing a web search for what this, or that assumes the creator is following some degree of a standardized naming structure. Although I've often run into people that just make up their own acronyms, and such, and for some reason assume everyone will know what they are. Then in some cases, their made up acronym is the same as another that already exists, or they just use the wrong one, like people that refer to "free to play", as FTP, rather than F2P, when FTP is an acronym for File Transfer Protocol.

Like for the DMD2, when I did a search here, it sure looked like it was a more porn related model, and even the images Stable_Yogi used for it were had more nudity than for other subversions, so it seemed to fit. I could only assume that was what was going on, when supposedly it refers to a faster image generation model. This is the problem, I can just guess what this, or that means, and my guess could be wrong. Which was why I was asking for at least some extra info. Thankfully, it seems Stable_Yogi agreed to provide more moving forward.

@Stable_Yogi 
If you decided you wanted to literally make 50 subversions, that's fine. I'm not saying you should make less, make all the ones you want to.

All I'm saying is I'd like at least a bit more clear info on how each differs. The info you relayed here on the general differences between each sounds good, tell us that, and that would be quite helpful.

Thank you!

I think the best way to get that kind of angle is to use the prompt:

(she is taking a selfie in her room, arm out, from above)

this gives the angle you're looking for 😁

(I've uploaded some examples in my gallery 👍)

USE SELFSHOT AND REACTOR THE FACE ON AFTER

I usually use selfies with weighting.