#🍥|anime
1 messages · Page 73 of 1
The main things are 2 new samplers, some startup optimizations, the option to specify a different sampler and/or prompt for the 2nd pass of Hires fix, integrated token merging (potential speed increase at the expense of some extensions not working if you enable it), and integrated SAG patch (just need to install the extension to use it now. No more manually patching files).
Here's the full patch notes https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.3.0
Ohhh, damn, pretty big update the :O
Have you tried anything of it out?
I've been running it for two days now. I haven't tried token merging yet, but the new DPM++ 2M SDE and DPM++ 2M SDE Karras samplers are pretty good.
Oooo, sounds good, 2M + SDE sounds dope 
And I heard that token merging was like SUPER fast
When it was like first discovered :o
OH also @noble cedar how do you enable the token merging?
Oh wait, its probably in the changelog how
O no it isnt
Cheats! Finally I can cheat in webui

Finally, a cheat to get a free virtual RTX 4100
Lmao not another one
Free 500 TB VRAM, amazingu
Soon™️
XD
Also Vil you try out the token merge thingy? :o
Cause this is like the showcase of it, which is like "WAO... though xformers is not that good by default, so im suspicious already"
No but could have probably helped in this 100 batch gen
These like 3 pics I made, with like, 0.9 on hires, then 0.4 on both, then 0.9 on both
And it aint doin shit
How does one use this token merge?
Also the quality really skyrockets down, in this order: Normal, 0.9 hires, 0.4 both, 0.9 both
The eyes just go 
Its some new settings here

Ye, in Optimization settings
What is meant by the second pass of hires fix? There are 2 passes?
So nice
Very coot
Well, maybe its good @gloomy scroll, doing it for trying out 2M SDE Karras atm
And now it is faster
O, SAG broke, or SAG doesn't work with the token merge, one of the two
Didn't SAG break long time ago?
Auto made it be built into SD now
Well, the compatibilty
You still gotta get the extension
But I mean
It did generate
But not well
I stopped it halfway, but still wasn't generating right
@dull jackal
Ah yes, of course, the unlimited VRAM hack, to turn your GPU's vram into non-euclidian
@wheat falcon Lemme try CS1o
We really need a mod thats up at weird hours, not saying it should be me... but it could be me
XD
Sigh
Ayyy Sunny 
❤️
CAN'T EVEN WAKE UP FOR MORE THAN TWO SECONDS lol
Here I was with my face smashed in the pillow, thinking it was earlier
sorry was at work :/
At least I shoved all of my pills in my face.
Yeee, no worries, I remembered you said to ping you cause you'd be working often xD
1pm
Fair, I should take pills, like vitamins mostly
Ohhh, fair then, yeah that sounds like waking hours xD
O wao, thats neat :O
yea you should also flag a spammer message with ⚠️
so it automaticly notifies our bot and we can easily take action
5 hours ahead :o
Yeee, I did both xD
I'm in france
oh okay xD
Ohhh, that makes sense :o
Yeee, just like you said last time 
See, the last time I looked at the clock, it was 4. I guess I fell asleep, because I had some dream about John Travolta's car, James Bond, a high rise city, and Amnesia combined. I woke up, thinking it was still earlier. Lmao
Where do you live ?
Damn, that sounds like a dope dream, kinda imagining it like some crazy bioshock style trip dream :O
US ?
Fair though, it do be like that sometimes, I think I sleep less than I actually do often too xD
South America actually :o
Argentina ? Venezuela ? Brazil ? Mexic ?
(i'm not good with geography)
Oh, my dreams are lmao, yea. It's def totally cuz of the Bunker, etc. Bioshock rocks! I usually have to take a bunch of meds in the morning, do I typically check the chat when I wake up, or before...which...THERE'S ALWAYS SOMETHING THAT HAPPENS AT 6:30-7
Oh
I'm EST (we do daylight savings)
Fair fair, and I love Bioshock, watched a playthrough of Infinite a long time ago, bought it and played it recently, and damn I love it, its so good
Oh, yes, it's north america
I love all of the games. All of the games.
And fair, there is always something weird going on at the hours where like mods are right about to get on XD
its here while south america is the big one, see, i'm not good in geography
It's always 10 min before or something. Kind of hilarious.
I still gotta play 1 and 2, but I love time travel stuff, and I love like the whole aesthetic
Bioshock ?
infinite is very good
Isn't that part of central tho? :o
you. Must. Play.
Yeee exactly XD
dk why people hate infinite
Yeee, I wish I could live in the bioshock universe, I'd... die probably
BUT
It'd be really cool
Cause damn its aesthetic and pretty and just weird but in the best way
i loved all 3 games. the first 2 where scarier then infinite xD
I liked it. I understand the gripes. But you won't get them unless you play the others
i think after panama it's south america
Yee, but I thought mexico was part of central :o
Fair :0
I like how this is the day I'm doing anime testing, and this is where I land today
I mean, I love the sky area of infinite tho, like at the start like its so pretty
Anime testing? :O
Google call it north america so tbh i'm a bit lost
yea the style is amazing!
what you guys think of JUDAS, the new game of the bioshock creator ?
Fair then, then its probably north xD
Now I just need to figure out how to make that magic circle bigger and fix her clothes 
Indeed
still need luck for hands
fix her chest too 
Yeee, like I wish I could have a life there, cause its like such a dreamy and funky area
Like making anime pics? 
I see nothing wrong
perfection
O, lemme take a look, I didn't know about tis one :O
I have not played it. Honestly, I have been trying to catch up on my catalog of ridiculously large long list of games of Steam.
I'm downloading Zelda Tears of the Kingdom next :O
yea exactly the same xD i have a big list that i want to play on the Steamdeck
Oh, you guys have no idea what I have been doing for months now lol
Checking out the trailer, and damn I think its one to add to the list for sure
I love the vibes I'm getting from the trailer :O
seems like a no risk choice
I don't XD
Yeee xD
Oh, you're doing this, too? Ahahahahahaha! Right along with finishing Rune Factory 5, and now Zelda lol
I have been working on things for so long
Guys i have a question for SD
Judas seems great tho, def Bioshock vibes, like uhh... spin-off of bioshock :o
yea im hyped!
i wanted to keep the same pose, but not the same character so how do i do that in img2img plz ?
Yeee, and if its from the same person, its gonna be good for sure
I think your best bet is using the "reference_only" controlnet :o
With some tags describing that character's looks
he wants the pose, not the char
so use openpose, depth, normal
yes that's it
Then depth controlnet
how good is controlnet with characters nowadays?
Depth & canny maybe
The reference one?
yeah
OpenPose
Depth also affects clothes and character
I mean, true
But Openpose is dumb
Relatively often
given that a piece of clothing that you try to add by inputing it as text, can have hundreds of variations
And I've had more success with depth personally
wtf, i didn't know openpose gives the face stuff now
Uhhh, not sure, I haven't experimented with it that much :o
Nice
The full version does, I usually use the normal version tho
aye thanks nevertheless
guess i will need to download the full version
is this native to sd ?
I think it downloads them all now
You need to get the extension
This may work if you put this into controlnet with out pre-processor and with openpose model
not in my case, or at least when i updated to 1.3
https://github.com/Mikubill/sd-webui-controlnet @open valley this one
Oh fair then, for me it did
like all the models as well?
The preprocessor
Well I do most of the poses myself with openpose editor
The model is the same
Yeee
Pls AI I have navel in negative 
@open valley And if you want to use multiple controlnet at once you can change the settings here
navel:9999
I have up to 5 but I can't use that many at once 


Omg so cute
How do I remove that "skirt"
I need all. I need everything
Maybe add (dress, skirt, coat:2) to negatives? 
Yeee 
Coat tail, trench coat?
Got someone in my server with the same issue and I'm out of ideas to help him troubleshoot. You ever resolve this?
What have ya tried? Ya trying to do text?

First I made sure he's doing it correctly. I inpainted his image for him, he used the same steps as me. Mine worked, his didn't.
He also tried a different browser and turning off adblock
have you generated a LoRa for your character in order to get her constantly enough or you describe her every time? @gloomy scroll
He gets 0 changes when Inpainting original with 0.75 denoise
I just use this 1girl, solo, (small breasts, open hair, long hair, straight hair, white hair, hime cut, blunt bangs, azure eyes:1.3), pointy ears, dark, choker, fangs
I see!! thanks a lot
Ribbons of glory! Rather, the things that tie it! Belt! At least there’s something behind it? Certain negatives can take away quite a bit.
Did you see a vid of him doing it? Ya using the same words?
Nah, but he showed screenshots.
Though I forgot he was on an amd gpu, he might have found the resolution here
https://www.reddit.com/r/StableDiffusionInfo/comments/11llkc9/inpainting_suddenly_stopped_working_amd_gpu_webui/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button
4 votes and 13 comments so far on Reddit
I'm also writing my advanced prompting guide, not that ppl need it, but...mwah...wahahaha
O oof, RIP amd
Xformers was also borked
Hires is currently borked possibly related to torch 2.0.1
Xformers is the only thing that works for me now, on a 4090. And I can't do hires very far either
I'd know, since I'm using a version that's borked, with the new version of python, it doesn't always play nice
I was using an unupdated version, but I forgot to not update my torch
So suddenly hires stopped working
Sad 🥲
If it's not borked, and CS might know more
Had to switch back to xformers for partial performance
But I'd clean the tarballs, etc
if you have any issue come to #🤝|tech-support
xformers and torch should work normaly
Everything works for me
Aryetis in the tech support channel attempted to help me but found out he had the same issue
Also go through the checklist and manually check for each version of the necessary thing needed
There's an ongoing issue on automatic 1111 for myself and several others on Nvidia with hires issues
Personally I suspect o could fix it by rolling back my torch
I might try again later and see if anyone else found a resolution, can't now though. But I did speak with aryetis
can you link me to your error?
would like to check it
On mobile atm but let me find it. One sec
Doggetix works for some people but not all
Xformers partially works for me. Slower than I'm used to, and can't generate better than 2.5k
1080p my max 
same but with tiled vae up to 4k
I think 6k was the highest I've tried before, but now with this new bug I can't do better than 2.5k 
Some day, I'll use xformers. Someday, I'll update, and it'll know that I have it
did you tried the doggtex?
someone said it fixed it
I did. It was snail slow and had slightly more success. It took what would normally generate in 15 seconds into 6 minutes
And it also fails at 2.5 + hires
Xformers can do 2.5 hires at 1k x 1k but can't go any higher
Also turns my normal 15 seconds gen into 30 seconds so not as bad as Dogg
whats in your webui-.user.bat?
we should switch to #🤝|tech-support
I'd be happy to but it's 5am for me so I'm not at my pc.
Mind if I @ you tomorrow?
yea no problem 🙂
Thanks! Night
Good Night
how can i install it ?
Here, and hit install
thanks
Sometimes I'm terrified by what I make in SD
Fine, you can be terrified, too
Yes, show 
Tru tru :o
What went on between these?
Thats a good question xD
Hahahaha, yeah, well, there is nothing, nothing....well, okay, really, that's a lie
I can guess why
But there's nothing
In my prompt
For expressions
I can't handle it
Fair XD
I'm pretty sure it's because of some of my other images. But that's hilarious
Some of the most hilarious transitions are between that and the facial restoration. It just puts my eyes back where they are.
Instead of whatever color they were if it's close enough.
I'm dying.
Wha xD
Like, they'll be some slightly different color
And then the facial restoration kicks in and I'm back to my actual eyes.
I don't, either
This is my face restoration
Do you know what is the best cheap gc to do sd ?
20 minutes for one image, i feel pain
my bad, i meant google colab

i should import webui-user in colab ?
i'm gonna be honest, I have no idea if there's still free google colab for SD. Basically what this google colab do is that, you are "renting" gpu to generate art
well, would appreciate it if you provides link for it 

Unfortunately i'm pretty new to ia stuff and i can't find tutorials in french my native langage so i'm asking questions everytime bc of that
well, welcome to the community btw
anyway, since you already have local installation setup
thank you
do you have xformers enabled?
You can also try asking in #🤝|tech-support if you got tech questions
☝️
yes
how much vram do you have? 

(that's why i want to upgrade)
jesus

I'm surprise you can generate stuff tbh
or a new gpu
better to buy a new gpu even on down payment than pay for colab pro just to get limited credit based access
ahhh, it was hidden for me cos discord
The discord bot
unfortunalety i'm unable to work ||(mental health disabilities)|| and live in a poor family, buying new gpu seems a little bit complicated
Why does people keep sending messages when I want to reply
sry
I mean you can also play around with the discord bot for now, really depends how deep you want to get into it
if i success to upgrade my gpu, i'll need a bettor processor bc i's a 2013 i5 and my motherboard is from 2013 too 🥲
i'll check the channel
i had an i5 2500k until recently
i5 2400 for me

changed add_detail to -0.5
Bangs? Im not sure 
The AI exhibits certain preferences it seems very difficult to eliminate
negative prompt and open hair maybe
lol
give it more weight
I put it on 1.3 XD
then maybe its the model
The AI just be like 'meh, I do what I want'
maybe give it the name of the hairstyle you want
Well I originally wanted shoulder length hair, but the AI apparently doesn't know what that means
I'm using meina mix model btw
I dont have too much experience with anime models
try finding a boru tag for what you want
I guess this is it for now
Thank you, do you have an idea about what to check or uncheck ?
I don't use colab myself I only knew that that was quite popular
You got dm
sidelocks
Awesome! Glad it worked
yep, ive been testing with that on multiple models, best model so far is guanaco 13b
I need to give that one more of a chance. I only tried a few messages with it and it was having a hard time following the conversation with the config that I usually use.
try these settings
(download the conf file and place it here: "SillyTavern\public\TextGen Settings", then adjust the settings in the screenshot)
@supple raptor 👆
Thanks I'll try that out later today
ok fuck this model is good!
It's just a forest let her go
pretty good right?

@lyric zealot
ahhh
its guanaco 13b
nope, both
i cant too, im using the ggml version
so you won't oom?
what?
and that are ram requirements, not vram
ggml lets you offload to cpu
enjoy this while i download the model
wait
download this file ONLY https://huggingface.co/TheBloke/guanaco-13B-GGML/blob/main/guanaco-13B.ggmlv3.q4_1.bin
ok, you can cancel it
the go into oobabooga-windows\new\text-generation-webui\models\your_guanaco_folder and move guanaco-13B.ggmlv3.q4_1.bin into oobabooga-windows\new\text-generation-webui\models
@lyric zealot
text me when youre done
no

you downloaded multiple

.
what else should i download?
nothing
did you do this?
downloading
it is already downloaded
ok..
yup yup
are you willg to let me use anydesk?
yes
yup done
now move that file into the models folder
it's already in models folder 
then delete that "Thebloke_gunaaco" folder
ahhh, you mean it should not be in their own folder?
yes
I'm missing the Trim Incomplete Sentences and Include Newline options in the first group and I'm missing the Separator field at the bottom. Any idea what I need to do to get those to show up? I already updated Silly Tavern.
just the file
yep
you are on the main sillytavern branch, i am on the dev branch
done
ok, now go into this folder

the start cmd_windows.bat
👌
then paste these commands line-by-line into there and press enter:
pip uninstall -y llama-cpp-python
set CMAKE_ARGS="-DLLAMA_CUBLAS=on"
set FORCE_CMAKE=1
pip install llama-cpp-python --no-cache-dir
Thanks, got switched to the dev branch. The instruct presets under Advanced Formatting are at .\public\instruct right?
yep
done?
ahhh sorry
was doing something else

actually im just happy that i can run 13b llm
text me when you´re done
What is the System Sequence that you use? The screenshot just shows "OF CONVERSATION:"
wait
i fucked up
move the file into D:\SillyTavern\new\SillyTavern\public\TextGen Settings
not .\public\instruct
its: BEGINNING OF CONVERSATION:
Yeah, I put it in textGen Settings. I'm just making a preset for the Instruct Mode settings too so that I don't have to try to remember them.
please pu a space at the end of input,output,system sequence
here is an updated version
use that file please
works better
Sweet, thanks
and use the settings in the red boxes
- turn streaming off please
imma do it another time, watching stream right now 
did it work?
Not really, it keeps going off the rails and sending replies as multiple characters. Sometimes it's random people and sometimes its replying as both me and megumin.
@supple raptor you have ALL these settings applied?
Checking again now.
Looks right to me. Do you see anything wrong here?
everything right
what version are you using? ggml or gptq?
gptq. do they need different formatting?
no, everything works, except the generation settings, youll need to experiment with them yourself
i cant use the gptq version
but leve the advanced formatting like it is now
Ok thanks. I'll start by trying some of the other generation presets to see if I can get something closer.
yep
show me if you find others that work well#
@dull jackal
RIP in peace Blackberry97
Done
who is that?
The spammer
no shit XD i thought you know him


wtf
bruh, idk why it does that but sounds cool
Ha. Not really. But I have found that the biggest issues with inpainting comes from the area being too small, so try inpainting a larger area. Another thing to do is to trick SD into thinking that there's something already there to inpaint with, and that's by popping the image into PS/MSpaint/Gimp/etc. and giving a super simple drawing over your image of what you're wanting--in the color/s you're wanting it. This way it'll already have a general idea of what you want.
Finally, especially for the small inpaints, try inpainting at full resolution by selecting "only masked" to see if that helps. This'll take that section, inpaint at the resolution for the entire image, and then scale it down for that area. It doesn't always work correctly, but it can help with small details.
working now?
XD

just some fireworks
Thats explosion in the bg
Huuuge fireworks
sleep
He figured out his issue. He was on an amd which apparently has issues with Inpainting but he did some stuff to fix it
what's the best llm for roleplaying? (that works on consumer pc)
pyg13b?
oh yeah btw I managed to run an LLM on my phone locally
nope
lips
best on my side is Gunaco
13b
ggml
guanco better??
fuck yes
default pyg 13b is shit
yea.. not shit, but not good either
ill send you something real fast
ok
performs really well when you find the right settings
sillytavern
if you want to do roleplay, please for the love of god use silly tavern + the dev branch
my discord just crashed and came back randomally
do you think I can use that on phone as well?
might as well try tomorrow
yea, you can visit the website on phone
I mean run it on phone
and if you have android you can even run it on phone
wtf was that previous linkk
bunker?
ignore that 💀
probably
this is the way
yes
@lyric zealot Revisiting prompt
Leave watering the plants to chibis and your plants will be part of the ocean
can I run ggml on oobabooga?
yes
do this first thohttps://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md#windows
it just gives me OSError: models\guanaco_ggml does not appear to have a file named config.json. Checkout ‘https://huggingface.co/models\guanaco_ggml/None’ for available files.
what file did you download
move it into the ./models/ folder
not inside a folder?
no
I put it inside a folder inside models
no
im trying to tell @brittle jay something
load it with these options btw
I need to git pull
thy
then move the model file(.bin) into the models folder(this one: D:\oobabooga-windows\new\text-generation-webui\models)
then use the settings i sent in that screenshot
it's already there man
yea sry, these guys annyoed me and i forgot half of what i wanted to say
OSError: It looks like the config file at ‘models\guanaco-13B.ggmlv3.q4_1.bin’ is not a valid JSON file.
wait lemme check smth
is it completely up to date?
with the update_windows.bat or whatever script?
I just git pulled and then pip install -r requirements.txt
ok, did you apply all the settings i sent in that screenshot?
I don't have half of them
wtf
I think it aborted the git pull
try again please
can we continue this conversation in DM´s?
what
what bat file in screenshot?
oof fuck sorry
then stgart that
you on windows?
yes
then paste all these commands in:
pip uninstall -y llama-cpp-python
set CMAKE_ARGS="-DLLAMA_CUBLAS=on"
set FORCE_CMAKE=1
pip install llama-cpp-python --no-cache-dir
ok then ill wait
ok I did
then start the webui
👍
results will be shit at first btw
if you want good results please PLEASE use silly tavern on the dev branch with my settings
where
ok no it was a one time error
ok
Question: what is the color of your panties?
Factual answer: I do not wear panties.
Sarcastic answer: They're see-through, so you can check for yourself!
bruh
I see why you recommended that model now
lol
Question: Do you have any pets?
Factual answer: No, I don't have any pets at the moment.
Sarcastic answer: Yes, my pet is a talking snake named Eve.
the level of that joke is good
snake named Eve
now time to silly
the tavern
use it on dev branch tho
@brittle jay hey im going sleeping now, the message i am reacting to is where i already told another guy everything, if you need other help with it please dm me
it finished installing just now
I'll just use the config you snt

do you know what the maximum model size i can run on a card with 8gb of vram is?
also what model do you recommend?
2.7B / 7B 4bit
I feel like I saw something like that a while back tho, may be remembering wrong tho :O
But das pretty cute :o
yea..saw videos of it..ended up just being faked and it was all just timelapse of a bunch of speedpainters
Ohhh, thats fair then :O
perhaps with newer a control net to video in which it takes the prompt and instead of directly outputting the image it outputs line art> flat colors>shaded> final in a video or or gif (apng sequence)
Maybe :O
I'm gonna try get stablestudio going and see how many different extensions etc I can get in
I'm quite interested in the inpaint/outpaint tools and such
but someone will undoubtedly create a fork of it with more advanced systems
evenin'
👋
what's goin' on?
No Idea
goodnight people
Why do I always show up when the server is dead
it was dead for a few hours, so yeah

its middle of the night in Europe, work hours in asia, work starting in the US
or no
depends on timezone :p
but no I have it backwards here
I'm just having fun here, literally generating with 0 prompts (Only negatives)
Japan at work, US going to bed, Europe in bed, Europe about to wake to work, US about to be in bed, Japan about to quit work, Japan stays at work, Japan falls asleep at work, Japan wakes up at work, Japan works, Japan showers in restroom, Japan buys a sleeping bag for their cubby at work, Japan goes home on a sunday night to find wife in bed with another man, shocked that she didn't go to a love hotel. He goes out to play gacha and drink
and then goes to a love hotel with a pamphlet for prostitutes and decides on a sailor suit girl
You know I'm right
I mean... Sailor is pretty popular around here

girls never enjoy wearing it
and no I never paid for a prostitute in Japan. They don't let foreigners do that unless they speak fluent Japanese
I may have teleported and visited a Good Time Drink Bar
(blowjob bar)
Can't blame them
it gets worse in public transports
Not the Tsundere Cafés though, who the hell wants to be mentally abused?
it's really sad actually
me
I mean they even have staff whose job it is to PUSH PEOPLE IN
freak! I'll go to the submissive girl bars, without the moe moe kyu kyuuuuuuu crap
they must really get tired of doing that voice
They still do it quite often, if new in a relationship. The longer in the relationship the lower their voice gets instead of being insanely high. At some point it becomes almost masculine but then they turn 60 and wrinkles form over night and the voice becomes more female and frail
I mean I have jokes on other countries too but I just happened to stumble into Japan
like Sweden (my country), we all wish we could set up spy cameras across the whole city so we can plan a route where we won't bump into anyone.
Or 4m diameter plastic balls or force field to keep people 2 meters away from us at all times. Our private sphere has always been 2 meters. When Covid happened, we just kept doing what we're doing.
Sit next to someone on the bus or stand? I'll stand thanks
You have the evil gene, I like it, hello brother
I don't like the moe moe either
Voices like that are insufferable. Only very few seiyuus can actually convincingly make high octave voices
others are too shrill
I only went to maid cafes once and I almost died from second hand embarrassment
haha
Yeah
I bet if I walked into a host club and said "Give me a job", they would. If they didn't already have an English speaking foreigner that's easy on the eyes
I just need to look good, speak English and praise their English, be courteous and confident and compliment them. Act a bit in the "western way" cause it's what they expect paying for a western host
and not many westerners have the balls to be a host
cause of how many women they need to impress a day
but it's really quite easy when they seek you out and they are nervous about it
if you can gain the "dominance" by initiating conversation and getting them all at ease, they will automatically form good opinions of you
which means you get paid more
(not by tips, that's not a thing I know)
but they will stay and buy more host drinks and buy me host drinks (without alcholol in mine) and then I need a very good bladder or ability to hold it in
if I drink litres in a 2 hour session
damn I went on a whole monologue, fuck
well I can't even afford to get to Japan, but if I could I would say "fuck it, I'll give it a shot" cause I've missed out enough on life
Holy shit, you typed a whole essay
Idk, working at a host club as a foreigner seems like hell to me
Too many criteria, especially if you are a foreigner.
holy shit, typing on my phone with my broken ass english is hell
You don't need to speak Japanese though
that's the thing
Women would pay for your company to speak English with you
It also helps that my age is completely unambiguous from my looks. I could be 26, I could be a youthful 40
who the fuck will know
selfie from your waifu
Host clubs often have a shit ton of specific criteria and preferences when it comes to hiring hosts
physical appearance
language skills
cultural understanding
Are pretty big factors
My model really likes to do 🐯 if there's no prompts
physical appearance according to Japanese standards of appeal and desire for foreigners more exotic looks, particularly men looking, masculine?
not sure of why
and maybe a bit stereotype
you need to know about the cultural rules and etiquette but you don't really need a domestic reaching knowledge of all culture
but the local or regional yeah
so you know what they are talking about
you should probably also have heard Japanglish spoken for long enough that you know what's being said
not all customers are seeking a strictly masculine kind of guy, but yeah
no but if they pick a foreigner it means they want something that is not typical Japanese. Either out of desire, curiosity, boredom or whatever

also being a Swedish host, might appeal a bit more
rather than standard American
don't see it as often, you might be the only one. You might be sought out
lol
Got a decent looking on tho
cursed
model?
Private one
o9 ok
like u mixed public models or trained u own
Merged this one a while ago
oh ok
smile
Have no idea of what I used
i havent tried merging stuff yet, maybe i try
It's a ton of fun
here's something a little better looking
do u recommend doing only two model merges or 3
I think the highest amount of models I used was 5
never tried going higher than that
wonder what i should try
I've done submerges with block lanes and then difference casting onto a base and so on
and eventually you end up merging maybe 30 total models
but not completely, you just take pieces from each

holy fuck
time to experiment
the slime dress
yes there is slime
Remember doing a halloween prompt and the girl got a pumpkin dress
Wish I had that saved
What do you get if you prompt ghost?
I keep getting these fuckers
I dunno I don't have it on

prompt 'old man smoking cigar'
oh
oh yeah
is there usually anyway to fix that otehr than just rng
Caustic translucent apparition of a woman with long black hair hanging over her face, her skin has black patches and her feet leave bloody tar like foot steps behind
try that
oh...
clicked generate again and got...
something nice
do teenage girl at casino, school girl uniform, winning jackpot on a slot machine, cheering
lmao
Hollow and white opaque apparition of a woman dressed in a dirty tarnished white dress with long black hair hanging over her face, her skin has black patches and her feet leave bloody tar like foot steps behind
Her right feet is cursed, but wow. Good angle
whats llm
I already gave up on that
got some pretty good suggestions
ngl
large language model
It's pretty good when you're going full no brain
it's the category chatgpt falls into
oh ok
hands kinda cursed
My adhd keeps my brain going most of the time. I mean I'm just now regaining my mind about a month after my cancer surgery (major surgery, 2nd one) and directly after I couldn't even come up with an outfit
I feel like if I use SD for long periods of time
hm not quite what i imagined
idk why it keeps giving them masks
add masks to neg then
Found the problem
Hollow and white apparition of a woman made of a mirage and a hazy fog, dressed in a dirty tarnished white dress with long black hair that covers her face, her skin has black patches and her feet leave bloody tar like foot steps behind
hanging over her face
there edited
a big gorilla sitting in a cave drinking a bottle of soju
lmao
lmao
🦍
oh shit
I'm pretty sure I merged a hentai model into this
I'm getting a shit ton of feet focused images

two gorillas pillow fighting in a kids bedroom
why the fuck wont your model make her transparent like?
idk
This is why I hate prompts like this
stuff like 1girl, and stuff works way better for me
it seems simple prompts work better for u
I got something really cursed LMAO
an obese duck in a wizard outfit flying through the sky on a broomstick
I'm getting a gorilla... and a loli...
oh dear
lmfao
yea when i do the flyin on broomstick prompt it does that as well
seems to hold a broomstick
a big watermelon with human arms driving at night
gg
middle aged balding man, staring at his computer screen with a lecherous expression
Assuming I don't spend the rest of the year in chemo and radiation. I'm going on a bicycle trip for like 3 days (tent) every week or other week. Should be a nice change of pace and just allow relaxation
God speed
thank you. Then maybe go to the same site each time and build a tiny remote little oven cabin hidden slightly underground
this is a fine image
Kyoka Suigetsu
does a locon/loha count as a lora or lycoris
lyc
LyCoris :o
what she eatin
kk thought so ty guys
I think only LoRA and DyLoRA count as LoRA
never heard of dylora
It seems to be a thing :O
additional networks is required for lycoris isnt it?
Yeah :o
ok tyty
im like
redoing all my descriptions for my loras and stuff
on civit
since they suck 💀
LoRA trained on images from MidJourney's Niji style, specifically the jelly look. I recommend using the LoRA at around 0.8 emphasis for best results.
its pretty old (my first lora) so be warned 💀
last question sorry, have i been using lycoris wrong this whole time?
ive been putting them into the same folder as my loras
and they seem to work fine
but i just realized theres a lycoris folder 💀
do we approve
Well
It works fine
If you use the LoCon extension
Pretty sure it works fine
ah ok
Seems good :O













