Has your finger improved to 6.9 already?

Best foot checkpoint

May i know the characters newly added in version 6 and 6.9? And are the ones that were available in version 5 all still present in the new versions?

JANKU 6.9 is the best feeling JANKU yet.

An absolute masterpiece, outstanding. The rate of 5 fingers/toes is exceptional high. The images look very clean and high quality. The eyes and the skin look amazing. The backgrounds are very detailed and clean. Thank you for your amazing work.

Well your model is pretty sweet! but 6.9 keeps on trying to add pots to it unless Lazynsfw is around for some reason.

From the display image, the skin of the character doesn't seem to have that greasy feeling anymore. It seems that there's no need to use Lora to optimize the skin.

Seems to change the physical proportions when used in conjunction with character LoRAs as well as the style slightly (v6). Did side by side comparisons with WAI-illustrious v14 and it likes to default to unnaturally large breasts/buttocks. I can't deny that the prompt accuracy and consistency is far better, as well as more natural poses/expressions. Visual detail seems around the same if not slightly better. Notably doesn't require the usage of tags such as masterpiece, best quality, etc. to obtain high detail images. The vast library of character/styles is very convenient, however I usually make my own character/style LoRAs anyways.

The Discord link is showing up as invalid for me...

Everything is great, but how to avoid so many additional lines between the eye and eyebrow?

Loved v 6.0 but it had one small downside for me. Model was very poor at 16x9 resolution (1344x768) giving low detail level and tons of artifacts. Is this fixed in 6.9?

Compared to version 6.9, the 6.0 model performs noticeably better. When I use 6.9, the characters’ shoulders become narrower, and the colors often appear overly saturated. The details of moist-looking skin are also not as visually pleasing as in versions 5.0 and 6.0. I hope these issues can be improved in future updates.

report a bug: this model output pure black image with xformers, while other models do correctly.
env: ComfyUI latest, pytorch 2.10.0+cu130, xformers 0.0.34

In versions 6.0 and 6.9, only sketch and monochrome images are generated, and even if I set negative prompts, these persist.

I use the same prompts in version 5.0 and get colorful results...

Can you post the character list elsewhere? Can't view the discord as I refuse to share my phone number with discord (and your verification requires that)

Seems like we can't use artist styles in v6.9 anymore? They work fine in v6.0 though

Really good checkpoint! It's very flexible and can be combined really well with Loras. Keep up the good work! :)

Highly recommended!

ds加入之后莫名奇妙被踢了，然后邀请链接无效

┭┮﹏┭┮

great at gen feet can and futa.

what controlnet do you use for this model?

Excellent checkpoint. All my results were good so far, but is it good for lora training?

Hey Janku or Janxd, I havent commented yet but I just wanted to say how much I love your work and your models, I have been using your v5.0 since last year (I have studied the crafts since september 2025) and it is my main model and I have created my own style within your model finally! keep it up. (sorry I havent commented before now, but I didnt wanna engage too much until I was confident in what I could do, and I still feel like a beginner.)

my workflow is probably the most ghetto workflow in existence lmao xD 15 lora nodes, not all are used all the time, depends on what I wanna achieve. my flow looks very primitive, but shit is powerful, I even managed to geometry lock a pose and put it on different characters without controlnet, which is probably nothing crazy, but to me it was a breakthrough lol. 69/69 RECOMMEND. super crisp anime aesthetic and highly modular. I will attempt my first loRA (style lora) on your model so I can free up some lora slots and see what happens.

Joined your discord and somehow got locked out of it after accepting a role (not the Ban me one), tried rejoining and wouldn't let me even though I did nothing wrong

The prompt adherence of this version is amazing! Great image output and lovely style!

My Main Model i use for genning. but mostly using 6.0 cause 6.9 made the art a lot of change.

Using the lazynsfw embedding in the negative prompt has no effect. What's the best way to use this model for SFW images?

Why I use this CKPT to make pics but the result were all black ??? Whers's the problem??? I just check all my loras and upscales but it still didn't work.

Most people seem to prefer v5 for its 2D anime. While v6 leaned more toward 3D, v6.9 captures the TV anime style perfectly. Prompts like 'anime screencap, anime coloring' yield great results at 1.0–1.2 weight, despite some slight issues with skin tones and physique.

I used to use WAI v15 for its vibrant colors, but the anatomy and finger issues were just too frustrating. v6.9 is definitely my top choice now, though v5 remains the best recommendation for anyone wanting a pure 2D look.

I've found that v6.9 struggles with style consistency when dealing with longer prompts compared to v6.0.

A major strength of v6.0 was its ability to hold onto an artist's signature style. For example, using the ireading tag: v6.0 perfectly captures that thick, oily, and painterly line art. In contrast, v6.9 tends to wash out these artistic features and reverts to a generic 2.5D look as the prompt gets more complex.

Interestingly, my testing shows this style loss happens suddenly rather than gradually. While short prompts restore the intended aesthetic, the style seems to "collapse" once the prompt exceeds a certain length, as if the model hits a semantic ceiling and defaults to its base style.

Ultimately, although v6.0 might not be as "stable" as the newer version, its superior artistic flair makes it my absolute favorite and most-used version.

To me, this series is the best—arguably THE best—at covering diverse anime styles on Civitai. I love it so much, and I truly hope this unique strength remains its core priority.

Is this a NoobAI fine-tune or an Illustrious one? Need to know if cLiPVision works since NoobAI has Clip trained

How do you make a prompt that makes nipple contours visible through fabric? If I type "nipples brush against a fabric" it generates nude breasts. I just want barely visible bumps. -_-