#🍥|anime
1 messages · Page 235 of 1
i never even really understood what vibe transfer was for or its potential applications
Been like a year since I used controlnet cx
To transfer the vibe of other images to the one you'll generate
👍
I mean thats basically it, a Fren I know uses it a lot
Just throws a bunch of images in there and gen stuff lmao
hey that sounds fun!
Hey guys
👋
same site u gave me ?
Your PC!
The one available is a different version
oh
noooo

I couldn't tell
Forge even more cx
I think for smol purposes forge will be maybe better

Until smol fine-tunes appear!
hey can u do a tekken character called "eliza" with SD3 ?


how do we get it to work tho? keeps comking out super low qual for me
what sampler/settings/etc?
still haven't figure that out yet
i guess default stable swarm settings with more steps and 4.5 cfg seems to work but anime doesn't come out very well, def needs a finetune.
it's kayoko from blue archive
What are those ears called?
Lmao
idk , this is an anime character
i just put her name


there`s smol in anime too 
Fair xD
soon™️
Is me :o
fenrys
In use it's basically a much, much better trained ipadapter. NAI claims that's not what it actually does though.
i don't know what ipadapter does either 😭
||♂️||
no way
smol fenrys
Gotta love dolphin shorts though
watch me while I work it out!
are you this buff?
You should lift. I began lifting last year and gained like 25 lbs
the power of friendship compels you
Very Grave of the Fireflies vibe
It's kind of unfortunate that this SD3 model does details and backgrounds so well since it'll probably get ruined with all the training that needs to be done to fix humanoid subjects
oh you give me ideas
150 images generated so far and only 2-3 are even kind of ok
eww
not to mention the hilarious censoring
it avoids generating feet like it's the plague
i didn't know they used children to make SD..
now, without characters in the image, the model does pretty good
hey thats not bad for sd
it follows the prompting pretty good
not bad at all
this one i said 'field of purple and pink flowers' instead of just 'field of flowers'
full prompt: photo of a field of purple and pink flowers at night with mountains in the background and a comet in the sky
not a complicated prompt, but the image it made is much nicer than previous models for depth and coherency
just put that in copilot to see the difference
that SD version is a bit more realistic imo
Looks closer to an actual photograph a person would actually take
omg the hands
Night looks so good 
Captain!
How is hires on sd3
yes
would like to know that also
I do believe them stairs need some work

aww kitty cat
Cat girl

It's you @nova remnant
Pink hair and flower hair accessory!
that is sweet
As we set off from the castle, a new adventure awaits us!
Meow dventure
But first we stop at our local tavern for a drink and some treats
And away we go!
You got really big at some point between the castle and leaving the pub 🤣
Big potion
What model?
Do any of you know if there's a way to log into multiple discord accounts on my phone
Or something that isn't a headache to be able to have multiple logged in
I know you can login on computer and phone at the same time
Only way is to use the app for one and a browser for the other
If you want them both online at the same time
I don’t know what you mean, hires fix or just generating at a higher res?
generated at default 1024x1024
anime girl in a field of purple and pink flowers at night with mountains in the background and a comet in the sky
same prompt but generated at 2048x2048
seems like it really doesn't use the extra area very well
Enable tiling
that's without a refiner/hires fix stage, just straight 2048x2048 generating
yeah, looks like tiling is the answer for the refiner stage though, this is a 1536x640 gen
this is hires x2 without tiling
Yeh, I had the same issue, but then I read https://old.reddit.com/r/StableDiffusion/comments/1de65iz/how_to_run_sd3medium_locally_right_now/
with tiling
Still not great though, but I guess we'll figure it out eventually.
anime girl on a beach at sunset with tall ships in the background and mountains in the distance and seagulls in the sky
I really like how it does the scenery and prompting it is very easy, it just sucks the people suck
and not just anime either, people in general are just garbage
it does cats fine though 😄
Try a cat from behind, or side view. It wouldn't do it for me no matter what I tried (though that was a photo)
i couldn't get angles to work with people either
That kind of works for me.
closeup portrait of an anime girl wearing an oversized black hoodie eating a burger, standing outside a rock club

it's got some problems, that's for sure..
cyberpunk city at night with neon lights and flying cars with huge monolithic buildings in the smoggy background, one large building with a sign that says "SD3"
for prompt adherence it's actually kind of amazing
added high drone shot to the end of the prompt
the scale of things isn't great
Foggy?
anime girl wearing a black shirt that says "PRIDE" in rainbow colors
first try
what's wrong with smoggy?
It me :o
Dunno, just another word to see if it does better or worse :O
I'd try SD3 but I'm not sold on it yet :o
but it did smoggy just fine
i could have done smog filled sky or something, but it wasnt needed
you who owns 3591395 loras, how could you be persuaded by a day one model
i'm only trying it because i'm already using swarm and it's was a simple update and load model
and i wanted to see what the performance was like because sdxl was insanely bad with performance at first
so far, sd3 is about half the speed of sdxl
~4s on a 4090
I mean I almost was convinced, but I saw that it got like the SD2 treatment so I was disuaded :o
what treatment
seems even worse than sd2 if you mean the censoring
censor or not trained on nsfw
both

well, at the very least it has basically no idea about anatomy
Yeee, it has a few issues :o
anime girl bare feet
pretty much everything is some deformed monstrosity, reminds me of the 1.5 base model
half the time it won't render anything below the knees, it actively abvoids feet at all unless you specifically prompt for them
which is weird to say the least
it's almost like the 2b model is just the backgrounds and scenery with some lighting/shadow knowledge, maybe the 8b model has all the people data 😄
If you want random stuff, it's good at that i guess, just nothing anyone actually uses SD for 😏
anime large cityscape on the back of a turtle flying through space
an illustration of a raven sitting atop a pile of bones with a flower in its beak
I mean yeh, I think it will only be good with finetunes :o
for booba cough cough
Comfy hurts my brain, and Swarm is aight, but Forge is cooler
Just a better one in general
True :o
pew
too bad they're nuking forge
Whos nuking Forge? :o
nyooooooo
is it because its using comfy
they say it's because a1111 has added all the important upgrades to the dev branch (except the actually important vram improvements)

I mean, it seems like they are doing like a separate thing from what I'm reading :o
well, instead of making a version of a1111 that is faster and more stable, they're dumping that and they'll just be making experimental features for a1111
i had to drop a1111/forge anyway because they both break with the way my system is setup now so i'm stuck with either swarmui or comfyui, swarm is pretty good though so i'm ok with it
Well, I can't do without my PAG, SAG, and Vectorscope, so I need my Forge
And A1111 is ass
and early support for new models and stuff is nice since they don't have a bloated crappy backend like a1111 to rewrite badly every time something changes
A1111 goofy lookin ass crashing on 832x1216
just make sure you keep a backup copy of forge and never try to update it, it'll work until you need it to do something new
lmao
no version past 1.5 worked for me on a1111
it just broke since then
yeah, they broke a few things in 1.6 that still don't work, my sd1.5 model doesn't work properly on newer versions of a1111 because of it
not that i use it any more
thats very good x3
grabby fox
Yeee
Forge is basically dead. It's diverging completely from A1111 which will break virtually all extensions, and he's focusing on experimental stuff. The dev recommended people switch back to A1111 dev branch.
Fair, I'll keep my good ol install then, since A1111 is stinky :o
you bully a1111 is good enough


no 
a1111 doesn't let me run sdxl hires 2x 2k resolution on my 2gb pentium intel gpu
bad
ok fine here
lora:dpo_sd_v15_lora_v1:.752.5D, retro 80's style anime, ultra detailed, extra sharp, masterpiece, absurdres, (1boy)chibi Anthro(Panda-SnowLeopard)Vivid Sleek designer suit, posing glamour photo, multi-tonal background, hands behind back, realistic, photo, movie, soft edges, multi-shaded colors, anime overtones, line art style, sharp focus, elegant detail, HDR, detailed shadows, cinamatic lighting, unreal render, soft colors, black outline style, elegant, very detailed Cel-Shaded With elegant shaded detail's, soft shadows, illustration
good luck
Evil and not smol 
I broke my wildcards somehow and don't understand how to fix them. When I generate several batches with hiresfix, the first batch comes out normal, while all others have same hiresfix prompt as the first one. This is repeated with both UmiUI and the regular wildcard extension. Has anyone else encountered this?
I know this its from a manga isekai genre
He's so...
hes sooo
HE'S SO HAWTTTT🙏🏻
He can fight!?
Goody goody fighting style activate
This is yona from akatsuki no yona yup
oh this ?
i thought he was talking about this image
this is fenrys from lv2

whats a good way to add just like a couple frames of movement
animatediff or deforum?
nice
I step away for a couple days and SD3 releases? sheesh
sounds like people aren't really happy with it
anyone tried it here?
nope
waiting for something that reduces vram requirements instead of increases or increases usage efficiency
I have, posted some stuff yesterday
Started here #🍥|anime message
I see I see
oh thats nioce
its has some, honestly, very weird issues with anatomy
not just hands but some positions break the model
but its amazing otherwise 
like it could be something settings wise and not the model itself, some people were having success with the apparently simple prompts that were breaking it just by having a different workflow on comfy something something
but idk
I can't use it locally 
what are the local requirements?
yeah looks like it's pretty good at landscapes and such
a tiny bit less than sdxl by default
if I can't use it in forge it's a non-starter 😛
forge is going away anyways
forge
it is?
ye ylalyiasviel guy will turn it into some experimental stuff
for a1111
so things will break if you update 
interesting
well, back to Auto I guess
too bad.. there were some QoL things I liked about forge
just img2img it
inpainting
Worked
I wish I've remembered about inpainting before picking the hard route of multiple img2img heh
There are so many options everywhere
i really like inpainting, so easy to just draw a blob over a bit of the image and have it generate just that
really only played with it in auto1111 though, haven't tried it in the other UIs
Sooner or later it will be time for me to try comfy
It scares me a bit since I've been on auto1111 and am currently using forge
i would suggest trying swarm, it has some nice easy UI features that should make it easier to 'convert' but it also has a full comfyui interface in it
I stuck with a1111 because of the easy "click for picture" without having to mess with workflows and stuff. I never had a problem with comfy, I just ended up working on workflows more than making pictures
swarm is the nice middle ground, I like the interface better than a1111 as well, but I don't use a lot of extensions or anything
looks better now too
I'm more drawn to opening krita, doing some sketches and running it back lmao
I have a little fun in that 
When you wake up from a nightmare where disfigured bodies are laying on the grass, calm down and remember that in SDXL, you can always lay on the grass, and everything will be fine.
using img2img how do i get an anime filter?
do you think god stays in heaven because he too fears what he has created?

controlnet probably has stuff that can do it
i cannot find a SDXL/pony compatible inpainting controlnet model for the life of me
thaks il look into it
I'd like to grab I mean buy your biggest potions 
🤨 sure

she grew up 
leather shawl ? dress ?
Hey this is fun 🙂
1girl, light blue hair, twintails, yellow eyes, ahoge, long hair, medium breasts, white shirt, crop top, long sleeves, denim shorts, socks, white socks, closed mouth, smile, one eye closed, white background,
Pure luck lol

which one of these sd3 models is the 'right' one? 😐
none of them for now

hello goblin shop
har har
just the regular one needed it seems... others more useful for training
can't get it to work on swarm anyways, though... new vanilla install and just crashes. Nice.
it only work on comfy atm
other one doesn't have support yet
also the whole license thing is just meh
the model prob won't get much support and expand by community
it works on swarm, too, which uses a comfy backend
cuz of that
just... not for me lol
one eye closed, v over eye would be the right way to tag it on danbooru. some models don't do it very well though.
the license is a non-issue for anyone just training models to release them publicly, and users who just want to make images to send on discord or whatever else. it only matters for people who want to be able to commercialize the model through a paid generation service, or selling their generations. even in the case of selling your own images, a creator license is probably plenty for the vast majority of use cases.
I see
I suspect we'll see models from the Animagine team, Kohaku, and eventually WD once they get some availability at their compute source (though WD-V (SDXL based) is still planed before they start on SD3).
Based on a message from one of the people at NAI, it seems very possible that they may also train it, though that obviously won't be publicly available if it does happen.
worked 😃
how do i use Doras in pony xl??
same as any other model. like other LoRAs they probably won't work very well unless they were trained on pony or one of it's variants though.
the thing is that i use forge ui, and they say i need to use a "pull request" to be able to use it
that means that DoRA is probably not supported in forge. a "pull request" is how you request to make a change to a github (or similar) repo.
Miku the Explorer
oh ok, how do i do that though?
Normally, unless you know how to write the code to make it work, you would just open a feature request under the Issues tab on github and include a link to the DoRA code so that other people who know how to write the code can implement it.
Unfortunately Forge is diverging entirely from A1111 and becoming an experimental repo and these sorts of requests are much less likely to be implemented. The dev recommended everyone to switch back to A1111 dev branch unless they want to stick around for testing.
i don't really feel like i could go back to A1111 because i benefit too much from the performance enhancements of forge
plus the forge couple extension is way more user friendly than regional prompter
Most of the performance enhancements in Forge are already in the A1111 dev branch. With the changes that will be made to forge, virtually all extensions will no longer work so staying on Forge isn't really an option unless you're fine with not updating anything or giving up any features that aren't built directly into it.
oh i see, is there any new performance boost for pony in a1111? last time i used it was painfully slow.
in dev branch, yes.
what is that? sorry if i sound dumb idk anything about github
muscle
branches are a way for the developer to keep separate versions of their repo available at the same time. the default branch that you see will usually be the "master" branch which should be the one that's most likely to just work when you install it. you can view other branches that are available by clicking the button that I circled and selecting it from the list. before attempting to change your local install over to the dev branch i would recommend learning about managing git remote repositories so that you will know how to switch back to the master repo if something goes wrong.
@wet elm
damn ty! If i change to this will i lose everything i have installed like extensions?
yeah, you'll want to reinstall your extensions. models, outputs, etc can all be moved, but moving extensions will likely cause more problems than it's worth.
iit's unfortunate they're using the same forge repo for this new experimental thing instead of just making a new repo, basically screwing anyone that updates without knowing
they should just let the existing one sit and start a new repo
it's mostly unfortunate for all of the extension devs who have nothing to do with Forge and have (mostly) never officially supported it who are going to be the ones getting most of the invalid issue requests when people update, realize their extensions are broken and go and complain to the extension dev.
yeah, that too. Plus all of the people posting issues on the forge repo because X doesn't work, or "updating broken everything"
it's just a dumb move, changing the entire scope of an existing repo shouldn't be a thing
to be fair it isn't that uncommon for forks to change scope and become their own thing, just typically not on ones that involve so many outside projects with a userbase who primarily is lucky if they know how to use git clone
This is from a brand new anime called Boys of Summer
that I just made up
in my head
Is it normal that I never get a perfect character? Whatever lora you use, something is always wrong, one eye doesn't come out quite right, or you have an extra finger, there is no way to make a 100 percent perfect image. Another question, the new stable diffusion 3 medium is can you run using webview? @vital raptor @steady grail
Friend for her
getting a perfect one-shot image from any ai is still very rare. sometimes you can get pretty close, but look close enough and there are always problems. the nice ones that you see posted on twitter or wherever you look at images have likely been at least inpainted, and often a significant amount of manual fixes have also been done.
thanks
this one is probably as close as I've ever gotten to a perfect image with no inpainting or manual edits. and that was on a paid service (NovelAI).
nioce
I got an absolute perfect one once of a guy meditating. But it was one of the realism checkpoints and I am not into those. I eventually deleted it, but it was indeed amazing
i live here now
would watch
I'm innocent
Donut queen says she doesn't believe you
Afro's SD3 Experimental (CLIP ) SDXL Merge + SD3-First Anime Lora by Getsc. Upscaled and eye corrected (very difficult illustration prompts to get eyes right, even in portrait). I really like the colors for a basic simple Lora. People should let go of the initial frustrations with SD3 and start focusing on the Loras and Checkpoints. The VAE, t5 etc, can make it better. I'm not using the artstation hacks or the stars on these.
evenin'
have a 1920 x 1080 wallpaper
need 3840x1600
its not letting upscale to that
hmmm
one sec
3840x 2160
hmm should make it more anime somehow
well thats not working. hmmms
changes ideas
Gundam Decepticon
moves on
no one taking the 3840x1600 challenge?
3840x2560
1600 is an odd number
does some math
@keen hollow 3840x1600
or what would you like @keen hollow ?
just whatevs.. I like the widescreens 🙂
oops tried hi-res fix 4x upscale and it said no and crashed. dangit
heres a good one
Hii, any good sd3 model yet?
It has been literally 3 days since the model came out lol
there aren't even any good trainers yet afaik.
only trainer so far is the official diffuser one
there's also this https://github.com/bghira/SimpleTuner
nice, but also Stable Diffusion 3, This model is very new and the current level of support for it in SimpleTuner is preliminary:
Like with every new model release I'd say give it a week or two and every tool should be updated :p
Where's Pony V7 ?!?!
xD




Pretty good one!
needs more cyborgs 😛
CUDA Stream Activated: False Traceback (most recent call last): File "F:\AI\stable-diffusion-webui-forge\launch.py", line 51, in <module> main() File "F:\AI\stable-diffusion-webui-forge\launch.py", line 47, in main start() File "F:\AI\stable-diffusion-webui-forge\modules\launch_utils.py", line 541, in start import webui File "F:\AI\stable-diffusion-webui-forge\webui.py", line 19, in <module> initialize.imports() File "F:\AI\stable-diffusion-webui-forge\modules\initialize.py", line 35, in imports import gradio # noqa: F401 File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\__init__.py", line 3, in <module> import gradio.components as components File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components\__init__.py", line 1, in <module> from gradio.components.annotated_image import AnnotatedImage File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components\annotated_image.py", line 13, in <module> from gradio.components.base import IOComponent, _Keywords File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components\base.py", line 20, in <module> from fastapi import UploadFile File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\__init__.py", line 7, in <module> from .applications import FastAPI as FastAPI File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\applications.py", line 15, in <module> from fastapi import routing File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\routing.py", line 22, in <module> from fastapi import params File "F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\params.py", line 4, in <module> from pydantic.fields import FieldInfo, Undefined ImportError: cannot import name 'Undefined' from 'pydantic.fields' (F:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\pydantic\fields.py) Press any key
whats dis?
I cant start forge
uninstalled everything and installed, hmm but not responding
might as well go back to A1111 I think they are fazing out forge regular and going to just a repo dev branch or some crap...sigh
which is stupid since it works so well.. why break it. grr
Did you update forge?
yesh
rip
ileisysviel is not working on forge anymore, its using it for experimental stuff
so it broke
f now what?
you can go back versions...
how? commit hash thing??
I don't know a thing about that, https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/801 this is the post
that tells you which version to go to, I think, 
yoshh, less roll back
Mines working so i'm gonna stay on this version of forge as long as I can.
but here
ahhhh it doesnt workk
@nova remnant is it possible to compress forge folder and send it to me? not sure it'll work like dat
ahh me need forge

trying to make a cyber city on a jungle planet
I don't have forge installed
oh speaking of which , check out: Dystopika: A city-building sandbox for futuristic cities. No goals, no management, just creativity and dark cozy vibes. (on steam)


And yeah it should work

but forge isnt abandoned guyz its just been a few months 😃
forge never died, just became something else
do you have forge?
yes i just havent updated it
i cant do that but why u cant just download an old version ?
how?
there is only 1version on github, and dats also broken
yesh this one, not working
hmm less stick to a1111 for now I guess
if i install a1111 dev branch, can i share my old models folder? i want to get the dev branch since its more optimized but idk where to start/what can i maintain
yes, models (checkpoints), vaes, loras, embeddings can all be kept and shared between UIs
Do you know if you can use autism or pony models in the new stable diffusion medium? because copying what I had before doesn't work, and of course then I can't use lora.
Pony is a model (kind of) like SDXL or SD3 (medium) so you can't use one with another.
if you're still having issues, try deleting the venv folder in forge and then start it
I don't know if this is the right place to ask. How do you use adetailer in img2img with lora and maintain the same style as the image generated in txt2img? It seems like the face becomes generic when used in img2img, even using the same Loras
adetailer is inpainting with the same sd model and loras that you're using for the generation. depending on what you mean by "generic face" i would guess 1 of 2 problems. 1. you have face fix turned on in WebUI which uses a very mediocre method of restoring faces which might be exasperated when using adetailer. 2. if you're talking about multiple faces in the same image becoming very similar to each other, then the problem is that you are inpainting them all the same prompt in which case you can use wildcards in the adetailer prompt to create variations in the expression or anything else about the face that you want.
forge is dead as far as any normal users are concerned the moment that an update diverges enough from A1111 to start breaking extensions. could be the impending a1111 update that brings the current dev branch changes to the master branch
yesh tried, hmm unfortunately nope, meh gonna stick to a1111 until forge comes out again hopefully
cute, cute, cute
Smol 
is there like a more consistent way to do fishnets?
i inpainted these ones but i can't ever get a consistent pattern
I can't either
patterns like that are notoriously hard for SD in general
SD3 is such a joke 😭 they want to charge people to use this crap
go use sdxl base then
lmao
It was crap just like that
i guess that was always the plan. But i think it's more about the finetuning where the license got a little confusing
do you do more than just anime ? why would you want to use sd3
text inpainting
you can try the SD3 branch in auto1111
i think it was just updated, so probably super broken
also did i read right before that the forge author is basically just gonna ruin it (by "it" i mean forge)
yup
hopefully someone forks it and continues it in the original direction
i don't really want to go back to a1111 after getting used to the performance enhancements and also forge couple
i'm glad i didn't empty my recycle bin after deleting my venv folder for A1111
it'll make moving back much less time consuming, won't have to wait on a download

i saw someone say the dev branch of a1111 already has the forge improvements but the comments on the forge announcement are saying otherwise
it sucks that i literally just got settled into forge and i'm already going to have to move back

they are
stoooooopid

slight horror
ugly eyes very horror
Nah, that's just her character design
Anyone know what the correct settings are for Animapencil XL? I looked in the civitai page but didn't see it there
Hi,Is there someone that use comfyui that explain me how to improve my image quality?
finally
ط
this is better i think
can SD1.5 loras be used on XL ? because i have some 1.5 loras shown when i use XL or it doesnt matter ?
no
not in a normal way at least
even pony (sdxl) loras don't really work on base sdxl models (properly)
i've heard of ways to smash things together with creative comfyui workflows, but realistically it doesn't work very well if at all
oh ok

@nova remnant
2 smol 1 big
I bring cupcake
I'm sure she loves cupcakes. and all sorts of sweets.
one for you too 
my cupcake give it back
Have you got SDXL stuff to work with inpaint, or is 1.5 inpaint still the way to go?
I just got that type of donut now, I'm giving it to my little sister
when you scroll down to images on civitai and like only 5 of them have anything to do with the subject of the lora
Ugly hair girl picture
Why do so many img2img outputs have these weird blotches rather than aesthetic outputs?
Not the lines as that's part of it, but I mean the weird spots rather than uniform color
Something to do with controlnet because it happens when I use controlnet
Specifically, with controlnet canny. When I use it with a depth controlnet, it makes a repetitive fine dotted grainy pattern
probably because controlnet is mostly designed for people, not solid walls
Is there any controlnet for SDXL that's good for walls and buildings?
Might try the crew in #sdxl as well
Thanks I asked that now over there
thats cute
only the fox ears came out a little poorly.
Don't grab ears!
All mine, yours too

@nova remnant
Booba
(i promoted for medium size
, this one really wanted to pop)

Latest style, dunno if to post it or not
evenin'
Ello Omnia :O
what's up @magic wave ?
oh my gosh...
when you generate 30 images and keep tweaking stuff...
can't figure out what's wrong
then realize you selected the wrong model
many pillows for me behind her

Cute

Draw me a logo text is the Nenjiang Education Bureau
dinner disappeared...
cat prob got it
really old skool
anyhow.. be back later
aww
just understand these girls are pirates...
to be smol or not to be
You can't put a sword in his paws! He's already cut off his own tail!
Oh!!! that's it!!?!? then everything is fine! 🤣
how “poor Yorick” has changed!
I found something I haven't been able to prompt in this pony-based model --- a pull-up bar.
If you can do a character doing pull-ups/chin-ups, congrats. 😛
Iv'e been working with this character and just can't get her to do it!
I can do weights...
one of the problems with pony models is that a pose (or clothing, expression, etc) might be style specific
rar
like, certain styles can't do the smug face, but the model can do it
something like a pull-up bar rarely works anyway
almost
She's a Cheshire Neko cat girl 😛
hmm maybe i'm going about this wrong
ponders
closer
😄
??
close as I can get
wtf!? I said climbing rope and it does this
accidental pullup?!
haha
Me
It's too hot, she needs to work harder

(I'm assuming she does but icy magic stuff)
danbooru only has like 83 images tagged for pullups, i'm going to guess the model doesn't have enough data to do anything with it
can randomly get close, but it's never correct
doing a pullup is close as I can get
and they lived happily ever after!
pushups instead
Yes, it's just a new fluffy bowling ball! homing!
there..
how athletic they are
tokyo has really gone down hill after the zombie flu
How do you make those turbans that have a long cloth thingy hanging out like a tail? I tried but it ends up looking like the kind that fortune tellers wear
prob do a search on different styles of turbine's



