#๐ฅ๏ฝanime
1 messages ยท Page 55 of 1
oh no... i hope i didn't figure out how the add difference works...
on 1 ๐
i stabilize this?! i got a real fucking good model!
woo looks great!
so i think i got the errors figured out, why they happend in this merged model
There's a lot to test though, changing one small thing can screw up several styles, compositions etc
i'm trying to get the model to merge in a way, so the merger becomes EXTREMELY good at natural language prompting
and ditch booru prompting altogether
๐
i haven't changed the prompt... just fucking around with model mergers
that's the best fun with models ๐
I don't think it's a good idea to abandon booru entirely
the base being alfahentai + darkgemini : best natural language models there are atm
it's very useful when you want to add small things
agreed. but that's something base SD can do too... you 'shouldn't' need booru for that
@cyan burrow u tell me... but this feels pretty damn good as a gen
(i removed the indian/dark skin tags)
I dunno, I don't really like those portrait style gens
where there's just a pose for the camera with a random background
so many models do that
i know my dude, but that's the actual prompt xD
what i'm searchign for, is better context awareness, and improved image quality on natural language prompts
Where did you intend to release it? On your page, my page, shared page?
the quality IS of pretty damn high standard ... this is 0 lora's... purely the model
lemme try again, increasing difficulty for the LLM
is this on the merge?
yeh
if anything, i think i'd want it to retain my (lineart preference) and then use your model comprehension
that would make a REALLY good model i think
2/2 gens
imma saving this one as darkalfaprototype โค๏ธ not done with this concept yet
Sure, the only conditions I stipulate on my model page is to give my model credit and link to it and that no money is made from it
but if the model is composed of just our models, I think making a new account and page that we share and release it on is better
well... there IS 3rd model involved
except the 3rd model is used to remove EVERYTHING I DON'T LIKE
I see
sorry aurora, but you represent everything i hate in my prompting style
because this. this is what i like โค๏ธ
@cyan burrow i'd first like to get this model over to you, have you play with it, and see what you feel about it
i feel you can make this work even better than me (edit: fuck, didn't see the 3rd arm, i was THAT in love with these wings)
sure
I can work from start too by trying to get your model merged, or at least parts of it
and then compare
or add difference
this is add difference: alfahentai + (darkgemini - aurora)
okay
I'm working on adding eyes, hair, cybernetic implants, magical specifics and such to my model. So you can add specific hairstyles, eyes and so on
well, that's very good
Then I'm going to create divergent models that focus more on a specific theme or genre, I dunno if I will start from scratch or what but if I have the time and energy I'm gonna do that. But otherwise it's "just" the details
mad thought, wouldn't such things be better for lora's? i feel details/styles/themes/genres should be relegated to loras, make the base models as encompassing as possible (says the dude who trained 0 things)
or embeddings
nah, lora's and embeddings will end up conflicting with each other
contaminating each other
but well, not every one have very extra out there
fair point. concept loras do tend to bite each other
contamination is good ๐ฅฐ
half true. if this contamination confuses your base model, you just get rubbish
one of the more annoying things for any model is specifying hairstyle or eye details
sometimes gives interesting results that have nothing to do with their themes XD
I want to make those things work much easier
well, that's the point of weighing them well
honestly, you shouldn't need multiple loras biting each other for that
but it's funny, man lol
that sounds like base model stuff
no weight will generate something it doesn't know you're asking for
Once I get eyes and hair in, I'll look at other things that are difficult to prompt for. I don't have much energy though and the energy I have I spend on rehab
So I dunno how long it'll take
man, take your time and don't run
expand your boundaries little by little, living your life in a comfortable way. who runs to overcame the hard times, will run for the rest of his life
aight, night crew, you guys take over โค๏ธ imma go to bed
night night my man
me too, good night
nighty night
Nice
hmm
evenin'
hey lotta, new ver of my model comes soon hehe
nice I gotta try it
More variety and understanding of details, as well as slightly improved hands
let's see some more examples
Here's today's Tifa.
it's looking great @teal pivot
Yes, I'm currently testing different prompts to make sure everything is fine
I like the way the model does hugs
try girl,samurai,(traditional_media,illustration:1.4),flowers in hair,kimono,wielding katana,normal_breasts,(oni, horns, detailed_eyes),fighting_pose ,(masterpiece,detailed,highres,cinematic,atmospheric:1.4), (hotify:0.6) ?
ok few mins pls
upscaled result or just generated enough?
nice backgrounds
Great!
anyone have tips for character lora training? My source images are all the same background, so the results output pretty much just that one room as the background for all pics
You have LORA trained to make such a background because it is repeated many times, use a dataset with a variety of backgrounds so it is not repeated, then LoRA will not remember it
are those live interactions with a bot?
Would it be better to just do transparent background then?
no u cant train with transparent images
You can try, but I don't think the result will be good
with transparent bg
ty
the best way to use a different background, but the same character
Her left hand ๐
It's AI generated what do you expect๐
you guys never cut a hole in your mattress?
nope
lol
never intentionally
I hate when I have an idea of something I want to generate, then I forget. ๐ฆ
smoooooo
experiment made with Parseq
She's small now
Loli? Sorry, loli?

Never wanted to be food more in my life
same tbh
wish this one didnt look as young, but it came out so good
That lighting mmmm
Did I just walk into a room of teenagers? :<
i tried to make them more adult looking, some came out more teen-ish
More or less my preferred age group
I sincerely hope you're not too old for that
Have my last ap test tmr wish me luck yall 
I will bite harder
teeth

im working on making her more insane, going for that"my life is in danger"feel
Look up the crazy expression lora on civitai
thats what im trying to use
Is it not working
it is v2 i think
oh i dont know how things work then
Honestly its been awhile, i might be wrong
But i havent really used it since i first put it up
Ah
Then I only trained it twice
And back when i was still learning to train stuff
So idk how well it works with things these days
Everyone else seems to think it works fine tho
The last person she bit started bleeding clearly
indeed
perhaps i need a blood lora
no blood, just some people might not like it
shes feral
Fnaf vibes
Running is an option
what if she is faster than you?
i cant get one of her running oof
here is another similar one instead before i have dinner
Yes, as you can see in the first video. Text by LLM and then image by SD
slightly less detailed version
playing with sweetmix

is it just me or are sd based anime models better than niji journe
looks nice ๐
hi ^^
i tried switching back to unipc again, now they made it so it can do upscaling, unipc my new best friend again ^^
haha, some models work well with it
unipc?
no man, you are not, you are still sleeping and this is just a dream
I wish 
It's already 11 am!
weren't you lagging 6h behind me? it's 11am HERE
hey guys
NOPE!
hello alina
I;m form EU duuh
dunno
yeah, 11am here too
๐คฃ
nah, I use photoshop. I;m way used to it
eh... ๐ฆ now if that right shoulder was ok ๐ฆ
After working more than 10 yrs with PS... its hard to switch
honestly i'm in love with sketchbook
The one with longer hair is still better!
Is it that nice?
lol, this prompt is ALL OVER THE PLACE
๐คฃ
it's very compatible with my drawing table
and pretty intuitive and fast with everything i do
oh yeah, with tablet for sure
I have Procreate on tablet for drawing
for editing, still prefer to do it on PC ๐
procreate, let me see
@ocean reef ur husbandos are calling you
dayuuum, 2nd one is nice!
this look super cool
I like blond hair too ๐
well, you need to spend for art
u have trained your lora
I recommend it for drawing ๐
this is 0 loras, pure prompt
how u are getting so similar faces
it's the model
luck.
๐ฎ oh, i was working on DG3
no wonder the concept grasping was this good again
oh really
dg3 works really well with my natural language prompts
back to alfahentai ... i REALLY need to find a way to get the crisp lines from my model ported over to dg3 ๐ฆ
keep working and you will achieve what you want!
@merry tendon ur model is pretty damn good on the natural language too
seriously?! i've having the windows opened, it's 13 degrees outside, and still my pc is warming my room more ๐ฎ
guys I am opening my Shopify store any way that I can make mockup direclty in stable diffusion for this design
I want 4-5 pictures of males, females wearing this tshirts
It's the same for me... xD
before opening the windows (opposite sides of my room) my room got 24 degrees, the heater has been off ever since i got the 3090 xD ... well my home is rated A+ when it comes to isolation so yeah... any heat i generate in here will not leave!
I can do that in canva but ๐ I have to get premium
You can edit it on any random photo, then do inpaint and mask the design (so it doesnt change) and then render human
Thats the only way I can think of
and I think Shopify has nice mockups too
i dread the summer already
i prefer winter. can't be cold enough
๐ฅ
โ๏ธ
NOOO, pperspective... why u do this. the dude is epic!
could you give me random promot of a man?
or boy
I want to test it
with negitve promots please
Behind him stretches a majestic landscape filled with rolling hills, grand castles, and sprawling cities bustling with life. His kingdom extends far beyond the horizon, encompassing countless subjects who revere his rule. He stands tall and regal, clad in gleaming armor adorned with precious gems, his right hand wields a sword befitting his status, his left raised in blessing or command. His gaze sweeps over his domain, eyes burning with pride and love for his people, as he ensures their prosperity and safety. Around him, the sun sets behind the mountains, casting its warm hues upon the land he governs so ably. Negative prompt: negative_hand, (low quality:1.3), (worst quality:1.3), (lowres:1.3), (error:1.3), (fault:1.3), (undefined:1.3), (crude:1.3), jagged, spiky, ugly hands, extra arm, extra hand, split arm, missing finger, extra finger, three fingers, four fingers, six fingers, (disfigured:1.3), (unclear:1.3), (indistinct:1.3), merged fingers, (bad anatomy,:1.3) misplaced hand, misplaced foot,, cleavage, sex, nsfw, ecchi, slutty, leotard, (text:1.3), (signature:1.3), (title:1.3), Steps: 50, Sampler: UniPC, CFG scale: 5, Seed: 2257729821, Size: 640x640, Model hash: 7d50d7e3bb, Model: xynaptix_vObsidian10, VAE: kl-f8-anime2, Denoising strength: 0.45, Clip skip: 1, Token merging merge attention: False, Hires upscale: 2, Hires steps: 25, Hires upscaler: 4x-UltraSharp
LOL
that's the prompt i'm running atm
and negtive ?
When I finish my work, I will try out your prompt ๐ I like to use others prompt and see result ๐
updated for full prompt
I always use my negatives ๐
dude came to slay the infidels
what's your negatives? (in prompt lol)
hahaha
like a lot ๐
strange
chief, got a prompt? i can't seem to get it from the png
oh i dont save it to png because the negative one is super long (i grabbed heaps from internet in the beginning)
ill dm you the prompt, but i use a lot of lora
i should probably fix it
9000 tokens? lol
something big
man lol
when i first started i just copy and pasted as many as i could find and made it a style
it works so i havent changed it
every one does things his way
but it's such a waste of time
it's hurts me at a personal level lol
i am here to approve this cyborg waifu ๐๐
im working on getting a version that is high res too
my pc would be crying if it had eyes
lets just hope the upscale doesnt mess it up
u using different mecha loras? if so, i find it hard to get high quality results with that sometimes
nice
my pc spent 10 hours for a 2048x2048 from 1024x1024
10 damn hours
lol
mine does this in abou 5-10 mins, but it doesnt want to
im trying to get some wings in there
this next one is promising
we have wings! ill now move on to a different waifu with mecha suit
no way 10 hours???? 
Look at dat!
i need a better gc
Casually flexing proper swords ๐
last one is rly nice ๐
thats it, im trying swords next
hehhe
lemme try different model ๐
and fck me, I've turned SD....
ugh
no work gonna be done today
Same. 
๐
best i have so far
i use the datail lora
ill turn it off to see
different seed but this is what i get
slightly less detail but still a lot
hahah
clyde would never net me lol
also we going back to pink hair waifu
left crazy expressions on lol
oh gosh
got them crazy eyes
more guys
beeznees, your prompt makes my images generations come out EXTREMELY THICC
pov, she loves you
like. REALLY. THICC
what can i say except youre welcome
how mahy models do yo uhave?
i have bought a disk for AI alone lol
Still a lot of space ๐
600gb of models and venvs mainly lol xD
I generate AI on 2TB disc too but I have games installed there too
Cyberpunk takes a lot of space....
i wonder why iit went with finnish... aiska something finnish (or somatic similarities?)
<- has no idea at all
u never know what AI gonna give you :d
LLM prompting so good tho ๐
you have one installed on your PC or online one?
on my pc ๐
created a "character" which is primed to give me such descriptions
dunno why the raven, but i like our new overlord!
dunno why wings anyway xD
aiska, don't distract me too much, i should be working too!
@steel arrow i think this thiccness might be bordering on illegal
thats a lot of thick
using nerfs negative_hands?
does it help a lot? xD
mm I should be home again soon. play some games and think on how to improve certain aspects of DG3.
yeah, it's pretty good, he really nailed it down that it doesn't change other stuff in the composition too much
and if it works, you'll have to run ALL your husbando models again!
@steady grail
wow, that's me, it captured my presence so well XD
dafaq is even this. @steel arrow i told LLM to try your name again, and in some uWu flavour. lol
Model: abyssorangemix3AOM3, dont like it >.<
UwU
there's a reason why i tried removing THAT from my model ๐
im trying it now and i dont like it either
however noting is too bad to stop thiswaifu
๐
I'm working with your prompt all the time
gimme howler if you want another, i don't mind throwing people idea's in that machine
I like new ideas ๐ but lemme try all models, gonna take some time xD
myb I should lower upscale ๐ค would render faster
but wouldnt be so detailed
Anyhentai....
you need alfahentai in your life ๐ (blatant promotion)
well, it's not as much a hentai model as what i use it for actually... lol
xD
ARGH,#1 GHOST HAND !!! otherwise so cute
๐
not a ghost hand? ๐ฎ
well, it is
or sb is hiding well
anypastelAnythingV45
I go with DPM++ 2M Karras, all the time, bcz why not
these are pretty good though
i do to, but it seems that unipc can handle transparant stuff better
im trying
get 2B lora ๐
uhh... i think i know where this is going to end...
sb sleeping on couch and another on floor
good idea
ARE YOU SURE SD?!
sb?
sb - somebody ๐
ah
sry got used to writing it short ๐
same as my tho(ugh)
anythingAnd_anythingAndEverything (d1d3799699)
if it weren't for eyes here... would be good ๐
i dunno... something in the composition or style i don't like... i dunno, something feels off
it's not great ofc, would have to work on it
should i do a high res of this one?
trick question im doing it anyway
xD
ุงููุง
cutting it close
AnythingV3V5, ugh I dont like it at all
even shinier
well dayum, thats a lot of steel ๐
๐
scary
totally different style now ๐ beenyouToon
anyway, i should probably go to bed
its very hard, pink mecha waifu is too good to stop
wait... what time is it where you are?
Sweet dreams! Make it mecha! ๐
just click on profile pic to see region! U lazy ass!
(i have no profile pictures enabled) XD
then on nickname!
you havent tested dark gemini 3 Aiska? shame
oceania... is still big XD
to part with, 2b about to hug you
Here's smol image
CamelliaMix_2.5D, half of them are fembois xD
looks like eleceed a bit, including the red eye outlines
is princess ๐
or a witch!
cetusMix
darkGeminiV3
uhh... #2 lady, ur dress is on fire.
#3 tho
fire is nth to her ๐
i suppose that's nothing? XD
downloading darkgemini but not AH ๐ฎ my heart is sadge
whyy?
is good model (my model lol)
xD
strange how your image output looks so much cleaner vs mine ๐ฆ
dont ask me why ๐
pretty wings
agreed
This model gave wings to all of them xD darkSushiMixMix
pretty damn nice art style. lol@wings
well, it's not called brightsushi, is it? XD
damn, why can't I get in the mood to play any games. I have a massive backlog that could keep me busy for like a year if I played 10 hours a day
disillusionmix_3
Which I normally can't...Not anymore ๐
same. SD crushed that mood hard for me
hello how can I get cartoon model?
nothing to do with SD
I'm sure there are plenty on civit, just have to scroll through
Maybe I need to fight through that initial game time before I become drawn into it, that's where it falls apart for me right now
All I find is anime or real
Gotta keep scrolling unless you know the name for one, that's just how it is. They do have a filter and tag search which can help
ugly face!
facebombmix, thought it won't work well with guys ๐ฎ
what's facebomb mix?
model
obvoiusly, but it's not guy centric?
nope
ghostmix_v12
damn, these are good ๐ฎ
yee, looks rly nice
need some work with eyes but still good
with close up its better
hassakuHentaiModel
lol ๐ looks nice
not bad ๐
but that's 2... what happened to the others XD
hahaha
and disfigured
holokuki_v2
2nd one ^^
doesn't feel very anime-ish
ye, more realistic
I think I'm gonna make a card for this weeks pow
aaaanddd... we're back to weird swords again xD
I've already stopped looking at it xD
newbie question here: do you guys just have like alot of different models ready to go? or is that all the same model? im still new to all this
I have a lot of models, testing out now which one I wanna get rid of...
After some renders everyone finds a few best ones for themselves
Yeah makes sense, ive been getting used to SD with CounterfeitV3, really happy with it for now. Got to test the many features of SD on the same model made it easier i guess
maybe ill explore more models down the line
i have too many models
mixProV4
mogglemix_v1
okey, I will post final draft with all anime models I have, gonna take some time ๐
took me a long time to get my model characters to mostly hold weapons correctly, regardless what it is. If it has a grip, they will try to grip it
i suppose the grip depend from object to object. would be worth training every single object,
at least the most common used
the thing my model focuses on is what a grip or handle is
and that's where the hand goes
how did you get that pose and the perfect hand holding a sword
Model?
alfahentai of course xD
Thanksโค๏ธ
sure it can
just as you describe landscape
people are so surprised at things I take for granted with my model
like holding swords, beautiful backgrounds, translucency
I'm dooonee
no alfahentai
(the hentai part comes from hassaku lol)
okaay! I will download it! xD
the line is sooo clear, which model is that?
alfahentai + add-detail:-1
when I use add-detail:-1, the line is also vanish

do you know any lora that help emphasis the line?
spoilered because underboob
Soonโข๏ธ
whuuuut are you doing?
and a question that I already asked but noone answer me, how do I embed the noise offset lora to a model to create a dark version 
Nothing rly seed input just suddenly started loading
somebody pls teach me ๐ญ
Supermerger extension can merge lora to model
supermerger causes drunk accidents xD
wao, that's a big help, tks โค๏ธ
What kind of model did you create this time?
nah, still the same, but i want to try to get the clean lines my model has somehow ported over to the concept awareness that dg3 has
you need more anti-nsfw in there it seems lol xD
and thx โค๏ธ
ah, yeah, this model might need that
all the same
I dont mind bare chest ๐
Now I have to do test with them for sentences prompting
all of them, all over again? ๐
you're not nearly oom-ing, right?
oom-ing?
out-of-memory-ing
inference becomes extremely slow once you get to the limit
but I do 2.2 upscale
batch size 1?
4 xD
... it might be faster to do batch size 2 in that case
yeah, that one is wicked good โค๏ธ
but now I can tell which one I can use for storytelling prompts
yeah, some models respond great, some just... not at all
yeah, now I saw that models which I've liked before are sh!t
and it would take ages to tst it on different samplers >.<
how so?
some models really like high steps, i know mine likes 35+
hahaha
what gpu do you have?
2080
that's not SO bad, is it?
lemme find full name
nah, it works well for me
but for AI it's sloow
NVIDIA Gigabyte GeForce RTX 2080 Gaming OC White 8GB GDDR6
For my rendering work it works well
I can render 4k img in a 2 min
or less, depends on file size ofc
4k = 2048x2048, right?
nope
whoops xD
4096 x 2615
in 2 minutes?!
yeah, I have nice rendering program ๐
for eg this
saved in jpg bcz of web browsing
no point in saving png
must consume other people's bandwidth!
program is well optimized, and I can view almost same preview before rendering it too
shhhh meditation!
i should too, before i run out of time again
but my new memory will be arriving shortly so...
^^ it loves high steps ๐ฆ a good and a bad thing
yeah, at that point only unipc gets stuff done reasonably on my model...
UniPc is weird tho, sometimes it lows low stopes, sometimes its a nightmare
I dont use it that much
it can't go with high cfg
it simply breaks down with high cfg, i'm running 5 cfg
I always go with 7 CFG
cool thing: there's no 'right' way to do all this though. being able to fuck around is part of the fun i suppose
I use dynamic thresholding at 12.5 mimic 7

my model with add-detail:-1 and ultrasharp4x upscale
not bad
oh my... i thought the darkalfa prototype was busted... but... this isn't even half bad
this one is much better ๐
how big did he say his... thing was?!
๐คฃ
looks f****cking good
i really like cell-shaded stuff, probably more than that 'soft shading' or hyperrealism shit going on
"Less is more" it works so well here
meet your god in VR
Nice! Another villain ๐
nice views ๐
wow....
thats awseome! ๐
#4. HOLY F that one is cool
it knows that im harnessing the essence of the digital world itself ๐ฎ who told him
these are all extremely cool ^^
dude these look epic, like a crew ready to hack a supercomputer and beat some guards up
Eface is FBI agent but sush, keep it a secret
shhh... now everyone know 
welp xD
detain me, i confess, i'm a rogue android trying to push the boundaries of digital art
thats me in the tech-support bubble
lol
being powered by near infinite amount of user-request threads? xD
exactly! ๐
why i can't see my pointer any more? why my monitor became red? why my voice on the mic sounds so stupid?
feed on our pain caused by our lame problems! became GoD!
have you tried to turn it off and on again?
๐
the second one looks shocked
i can turn off my pc? because it's the 3ยฐ i change after the last black out lol
the screen became black i don't know what to do
my god, such lvl incompetence would kill me with a stroke lol
haha for sure ๐
"don't attribute to malice what can be attributed to incompetence"
lol
i messed with a customer last week:
Me: yea then restart the PC
She: Ehm, i dont know where, ive never needed to do that"
oh... the pain!
yea then she turned off her monitor and said it wont work all black
i should have just hung up
humanity can be so sad. so lucky i'm an android... with flash and bones....
and blood..
lucky you ๐
ehehehe
i just noticed how some models are much more closer to the original, using control net and canny, than others
some does an extremely good job in fueling the gaps, than others
i remember there being an upscaler with 200k in the name or smth that i wanted to try
but i cant find it rn
probably because it was a junk
victorian androids!
2nd one has rly nice bg ๐
my anime model can do better hands, 100 billion dollar company BTW ๐
that's the price to have more flexibility
that's the usual paradigm. do you want more accurate forms, but less variety? or more variety, with more mutations?
hello does it effect ? should I put it 40 or 60 for more good quality ?
depends on the sampler
if you use DPM++ 2M Karras, it resolves at ~15 steps
so you probably won't get any improvements by going higher
most people use 20 steps to be safe
i usually test a seed
with karras, max 50
and i still get just useless details
but i always start at 30
reminds me i need to get pow started
hahaha I;m doing it, plying with anime model now
Have one card I like but...
gonna play more
I was thinking of a theme first, unsure if i gonna go with it and gonna take the day to think about it some more
why this quality ?
yeah, theme is important, thats why I have a few different, even in styles
which doesnt make it easier to choose
we could make a card game with sd lol
easily ๐
these are my settings
no it does not ๐
I will make a yugioh game but its only pendulum summons 
you are evil

guys could you help ?
in what?
I can't get high qualit images
I got this in start
try with some of this prompts ((best quality)), ((ultra-detailed)), (official art), Extremely detailed CG unity 8k wallpaper, masterpiece, best quality,
wallpaper, extremely detailed, cinematic lighting, Realistic, highres, detailed, , an extremely detailed cg unity ultra hd, photorealism, ultrarealistic, hdr, 12bits,
I I didn't get a good image I won't upscale it
if I*
use them with inpainting maintaining 75 tokens
so it's time for die with slowing ?
cuz if I made it high resolation I need to wait
see how they influence you model
I render in those setting and 1 img take 1:45
3.5 it's pretty high
I use img2img
it works with img2img too
1600x1600, right?
1600x2250
what is this arm
use controlnet for better hands
yep
unfortunately it doesnt work all the time as we want it to ๐
I rarely use them, so wont help here ๐
and for controlnet you choose openpose_hand
and then correct model
check YT how to use controlnet correctly
What is this calc extension?
sd-webui-aspect-ratio-helper

2024 sad
is this a boy face ? and is this a correct hands ?
any tips for getting better hands? embeddings dont seem to work at all...
luck
what negative embeddings do y'all use?
face is fine for me
only way to fix hands is controlnet, if you want a high quality hand
how to download controlnet models ?
hmm, my tarot cards are a bit dull atm ๐ฆ
whats that
I have different ones to submit, just playing with models now
which one of the 5 million ? XD
don't tell me you fell for it's charm ^^
imgs are nice, wo why not ๐
that they are โค๏ธ
nice
however, i don't think a waifu can win the competition...
how to do like this lineart ?
add-detail lora
download link please ?
thanks


