exactly a reason why it going to be less popular than it could

I'm just curious... do you have some tricks that you use to get good results in aspect ratios like that? I have terrible results when trying to do 16:9... are there positive and negative prompts that help?

@Jenika 2 things - just woke up, not good at writing:
1. ADetailer
2. My neg prompts are quite long...
```
poorly_detailed, jpeg_artifacts, worst_quality, bad_quality, low quality, poorly_detailed, signature, patreon, twitter, username, watermark, logo, brand, sketch, poorly drawn, white capelet, bar censor, censored, mosaic_censoring, (deformed, distorted, disfigured:1.3), bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, long neck, blurry, obese, rubenesque, rubens figure, long body, long neck, long legs, wide lens distortion, unrealistic proportions, stretched limbs, elongated body, distorted perspective, wrong scale, breast scar, vertical line on breast, border, frame, egirl pose, e-girl pose, egirl face, e-girl face, subtitle, urine, vomit, menstruation, diapers , femboy, futa, minor, kid, immature, child, chibi, loli, lolicon, shota, shotacon, toddler, faceless man, faceless, shorstack
```

@Sephrajin Thank you for replying! I have a lot of those same things in my negative prompts... I end up getting weird stuff like, if my subject is reclining or laying down, taking up more space horizontally, I get things like boobs showing up in the thighs.

Oh, by the way... you have things like "(deformed, distorted, disfigured:1.3)" in your negative prompt, but those won't work that way. Negative prompts can't use weights like ":1.3", instead you should use (((deformed))), (((distorted))), (((disfigured))), or (((deformed, distorted, disfigured))). It's basically the same way as writing ":1.3"

@Jenika Thanks, will try the ((())) vs the (:1.3).
Do you get the deformations during first-gen or during img2img/upscale (which is where i do - but not always - get these things...)?

Other than that, i'd need to see a prompt of yours where you'd get deformations constantly.

And please, understand me right, from about 100 (local) gens, only like half of is not too boring and/or has deformations (granted, boring is more often than deformations, not talking about fingers though :D).

But the likes you mention, i hardly ever see - if any at all - on first-"draft"-gen thaht is.

If you do get those during img2img/1st-upscale (my workflow, article), i'd recomend lowering CFG to 0.05, helps a lot for me (most often re-/moved naval ; in certain poses).

EDIT:
You inspired me to do a tiny little update on my workflow article.
Hope this helps: https://civitai.com/articles/14013

@Sephrajin this is one of the few examples of it I could find, because I usually delete these bad ones as soon as i see them, this is a 16:10 ratio picture... that's the ratio for things like phones and tablets, and i don't think it's a ratio that is really trained into sdxl... https://gyazo.com/1cb000d32fb968b1d06af961f6ba1df1 and sorry it's just a screenshot, i had already converted this one to a jpg, so there was no png info left with the prompts and stuff.

But what i did was generate the picture first using an Illustrious checkpoint, which was perfect deliberate, i think. Then I do highres fix with a pony checkpoint because that's what I trained my own lora with and i denoise at like 0.15 there. I think my deformations like that happen during the highres fix, and if use a higher denoising, it happens more often.

I use this same negative prompt for just about everything though... "ILXLneg, ((worst quality)), ((low quality)), ((normal quality)), bad anatomy, bad proportions, poorly drawn face, poorly drawn hands, missing fingers, extra limbs, blurry, pixelated, distorted, lowres, jpeg artifacts, watermark, signature, ((text)), (((deformed))), (((bad hands))), overexposed, underexposed, censored, mutated, extra fingers, cloned face, bad eyes"

I use that negative prompt in my highres fix too though, and that could be part of the problem since it has a negative embedding for illustrious, but i'm doing hires with a pony checkpoint... I should probably fix that and try again!

@Jenika high-res fix... so, that is done during the very same "run" as the image itself. This is why i do "drafts" at 14 steps, then only those images i like, i'll do - as/with a 2nd step - an img2img upscale, incl adetailer *(and that usualy has no negative prompt - unless adetailer adds stuff i dont want - to face or hands)*, this allows for "rerunning" (aka lower denoise) the same draft image with different settings.

See my previously linked article for details.

-> As already said, here and in article - if and where these effects appear, i can only avoid them by going for an img2img/upscale/adetailer step witha denoise of 0.05.

If you've a spare 9011-ish buzz, of course!