Style of the artist <a target="_blank" rel="ugc" href="https://rule34.xxx/index.php?page=post&amp;s=list&amp;tags=devmgf+">Devmgf</a> <a target="_blank" rel="ugc" href="https://tensor.art/models/750843734826152978/Devmgf-+-Chibi-style-(Pony-XL)-v0.91">Tensor.art</a> | <a target="_blank" rel="ugc" href="https://yodayo.com/models/a8a40ebf-b117-427b-942f-d88b5ad17693/?modelversion=bb3f0afc-ee5f-432e-a1ef-a0870861665a">Yodayo</a>I recommend reading the description and comparison of the two different versions.Illustrious/NAI XLV 0.94 NAIsame as 0.93 but seems less overcooked.Trigger word: dvmgRecommend cfg scale: 3-5 (4)at least for <a target="_blank" rel="ugc" href="https://civarchive.com/models/926443?modelVersionId=1061268">this noobai xl model.</a>UPD. For NTR MIX XII Recommended weight: 0.7-1.3 (0.9) for original style 1Recommend cfg scale: 4-5.8 (5)Difference between v0.9 &amp; v0.91The key difference between versions is the dataset 1259 img in v0.9 vs. 830 img in v0.91 Lineart and sketch are better at 0.9, also skin color and muscles seem a bit more original and stable here. Nevertheless in v0.91 the style is more recognizable, chibi is more stable, also almost all caption are removed here, but I think the flexibility is not affected.What to choose? I'd suggest testing both, as I've also given some comparisons and a great many images below. The differences in standard use are not so significant. But for chibi you're better off using v0.91, and for a more massive or anatomically correct body, v0.9 is a bit better.All images have a prompt you can look at and choose what you are interested in.For all PDXL versions: Minimum recommended steps: 28 Works with all source, feral have fewer changes. Recommended to use adetailer, hiresfix.v0.91 Trigger word: dvmg For chibi: dvchi Recommended weight: 0.8-1.2 (0.95) For more original style: 1The style is quite similar when the amount of prompt is reducedPrompt recommendations:Recommended words for original style: anime screencap, flat colorsIt rather simplifies the style, the changes are not as strong.Words with more impact: shortstack, cute, muscularAlso: chibiWorks with either of the two trigger words, but is much more stable with the second one.Also: white pupilsWith that word, I recommend using in a negative prompt: heart-shaped pupils Possible problems/Non-obvious points: The strength of the style can vary greatly depending on the prompt, as well as theoretically the default background is more likely to be simple.v0.9 Trigger word: dvmg For chibi: dvmg_chiNote that the chibi trig word has been changed! Recommended weight: 0.7-1.1 (0.9) For more original style: 1Prompt recommendations the same.Personally, I like both versions equally, which is why I posted both.

Devmgf + Chibi style (Pony XL, NAI/Illustrious)

Recommended weight: 0.9Recommended in case of deeper or flexible diving.

devmgfPDXLv0.9tx1.safetensors

v0.9

v0.94 NAI 1.1

v0.93 experiment NAI

v0.91

Style of the artist <a href="https://rule34.xxx/index.php?page=post&amp;s=list&amp;tags=devmgf+" rel="ugc noopener noreferrer nofollow" target="_blank">Devmgf</a>I recommend reading the description and comparison of the two different versions.Difference between v0.9 &amp; v0.91The key difference between versions is the dataset 1259 img in v0.9 vs. 830 img in v0.91 Lineart and sketch are better at 0.9, also skin color and muscles seem a bit more original and stable here. Nevertheless in v0.91 the style is more recognizable, chibi is more stable, also almost all signatures are removed here, but I think the flexibility is not affected.What to choose? I'd suggest testing both, as I've also given some comparisons and a great many images below. The differences in standard use are not so significant. But for chibi you're better off using v0.91, and for a more massive or anatomically correct body, v0.9 is a bit better.All images have a prompt you can look at and choose what you are interested in.For all PDXL versions: Minimum recommended steps: 28 Works with all source, feral have fewer changes. Recommended to use adetailer, hiresfix.v0.91 Trigger word: dvmg For chibi: dvchi Recommended weight: 0.8-1.2 (0.95) For more original style: 1The style is quite similar when the amount of prompt is reducedPrompt recommendations:Recommended words for original style: anime screencap, flat colorsIt rather simplifies the style, the changes are not as strong.Words with more impact: shortstack, cute, muscularAlso: chibiWorks with either of the two trigger words, but is much more stable with the second one.Also: white pupilsWith that word, I recommend using in a negative promt: heart-shaped pupils Possible problems/Non-obvious points: The strength of the style can vary greatly depending on the prompt, as well as theoretically the default background is more likely to be simple.v0.9 Trigger word: dvmg For chibi: dvmg_chiNote that the chibi trig word has been changed! Recommended weight: 0.7-1.1 (0.9) For more original style: 1Prompt recommendations the same.Personally, I like both versions equally, which is why I posted both.

Devmgf + Chibi style (Pony XL)

Recommended weight: 0.9

Recommended in case of deeper or flexible diving.

Files with SHA256: da0319209aad720ae17e88b425e04e30f468905c2a9f04583dba414e15d0eb11