/ldg/ - Local Diffusion General
Anonymous 01/15/25(Wed)01:06:26 | 317 comments | 71 images | 🔒 Locked
I am a top 1 percent poster and I certify this thread SLOP
Discussion of Free and Open-Source Diffusion models.

i2v Waiting Room Edition

Last Bread: >>103893498

Newfrien Guide: https://rentry.org/ldg-lazy-getting-started-guide

>UI
Metastable: https://metastable.studio
SwarmUI: https://github.com/mcmonkeyprojects/SwarmUI
Forge: https://github.com/lllyasviel/stable-diffusion-webui-forge
reForge: https://github.com/Panchovix/stable-diffusion-webui-reForge
ComfyUI: https://github.com/comfyanonymous/ComfyUI
InvokeAI: https://github.com/invoke-ai/InvokeAI

>Models, LoRAs, & Upscalers
https://civitai.com
https://tensor.art/
https://openmodeldb.info

>Training
https://github.com/Nerogar/OneTrainer
https://github.com/kohya-ss/sd-scripts/tree/sd3
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts

>HunyuanVideo
Comfy: https://github.com/kijai/ComfyUI-HunyuanVideoWrapper/
Windows: https://rentry.org/crhcqq54
Training: https://github.com/kohya-ss/musubi-tuner | https://github.com/tdrussell/diffusion-pipe

>Flux
Forge: https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050
ComfyUI: https://comfyanonymous.github.io/ComfyUI_examples/flux
DeDistilled: https://huggingface.co/TheYuriLover/flux-dev-de-distill-GGUF/tree/main
Pixelwave: https://civitai.com/models/141592/pixelwave

>Misc
Share Metadata: https://catbox.moe | https://litterbox.catbox.moe/
Img2Prompt: https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-two
Archive: https://rentry.org/sdg-link
Samplers: https://stable-diffusion-art.com/samplers/
Digital Art Software: https://krita.org/en/
Txt2Img Plugin: https://kritaaidiffusion.com/
Collagebaker: https://www.befunky.com/create/collage/
Video Collagebaker: https://kdenlive.org/en/

>Related
>>>/aco/csdg
>>>/b/degen
>>>/c/kdg
>>>/d/ddg
>>>/e/edg
>>>/h/hdg
>>>/trash/slop
>>>/vt/vtai

>Local Sexting
>>>/g/lmg

>Maintain thread quality
https://rentry.org/debo
Anonymous 01/15/25(Wed)01:20:44 No.103900444
https://www.reddit.com/r/StableDiffusion/comments/1i1q907/hunyuan_videos_naturally_loop_with_length_of_201/
Chat is it true?
Anonymous 01/15/25(Wed)01:31:57 No.103900502
fa495f69-87d7-4aae-86ca-66b355e5a908
Is there a way I can generate pictures like this on my 4070ti? Where do I get started?
Anonymous 01/15/25(Wed)01:51:54 No.103900615
>>103900502
>Is there a way I can generate pictures like this on my 4070ti?
yes
>Where do I get started?
not telling you
Anonymous 01/15/25(Wed)01:55:37 No.103900638
>>103900615
>not telling you
not very nice of you anon
Anonymous 01/15/25(Wed)01:55:41 No.103900639
>>103900444
>201 frames
that's a lot...
Anonymous 01/15/25(Wed)01:56:03 No.103900643
Blessed thread of frenship
Anonymous 01/15/25(Wed)02:10:30 No.103900720
Comfy said that going for low_vram mode was equivalent to the block swap but that's complete bullshit, it loads the model completly to the gpu because it assumes it's enough space, which is true at the begining, for example it loads at 22 gb, but once it starts doing its steps things the vram usage increases to the point it overflows my 24gb vram and my computer freezes, c'mon comfy make a fucking manual block swap node like kijai and stop being a fucking stubborn motherfucker, god I hate those fucking autists
Anonymous 01/15/25(Wed)02:12:50 No.103900733
ComfyUI_03590_
Anonymous 01/15/25(Wed)02:13:55 No.103900745
ComfyUI_03729_
Anonymous 01/15/25(Wed)02:20:42 No.103900779
>>103900720
>Comfy said
>but that's complete bullshit
What did you expect?
Anonymous 01/15/25(Wed)02:31:54 No.103900858
>>103900444
holy shit it's true, I did a 201 frame gen and it looped back to the first frame, this is cool
Anonymous 01/15/25(Wed)02:53:57 No.103901022
>>103900502
The 4070Ti is a great card if you're doing anything that doesn't need a lot of vram. It can do the nf4 version of flux comfortably and very fast, but if you want LoRAs and things you're out of luck unless you're ok with genning at low resolutions.

In your situation you'll probably be genning the image at maybe 0.5-0.6 megapixels so that you can use a LoRA to get that style of image, then upscaling it and going over it with flux-dev-nf4. That's not ideal.

If you're using something SDXL-based then the card is more than adequate.
Anonymous 01/15/25(Wed)02:55:58 No.103901034
>>103901022
>megapixels
feels like this is a way of talking around pathetic base resolutions to scam potential VCs
Anonymous 01/15/25(Wed)02:58:51 No.103901054
>>103901034
nta but 0.5 megapixels was the max first pass resolution until only mid 2023
Anonymous 01/15/25(Wed)03:00:53 No.103901063
>>103900444
yeah it worked for me.
Anonymous 01/15/25(Wed)03:14:51 No.103901149
>>103901022
thanks anon, I'll try to figure it out.
I can always buy a 5090 if I get into it.
Anonymous 01/15/25(Wed)03:22:38 No.103901198
so, did illustrious render controlnets useless?
No? Didnt think so.
why THE FUCK would i use illustrious cus muh prompt adherence when the tradeoff is that it LOOKS LIKE SHIT unless you some wizard when i can use pony with controlnet and get perfect anatomy and perfect artstyle out of the box uh??
Anonymous 01/15/25(Wed)03:24:46 No.103901211
????????????????????????????????????????????????????????????????????????????????
Anonymous 01/15/25(Wed)03:25:01 No.103901214
>>103901198
>muh prompt adherence
and compared to flux, the prompt adherence is still shit lol
Anonymous 01/15/25(Wed)03:27:04 No.103901224
are you people bots
Anonymous 01/15/25(Wed)03:27:24 No.103901228
>>103901214
haha xD totally sis
>pony = chads
>illustrious = retarded newfags who like to suck thick and veiny cocks
Anonymous 01/15/25(Wed)03:28:26 No.103901243
>>103901224
If you mean autistic, yes.
Anonymous 01/15/25(Wed)03:28:59 No.103901247
Can you VRAM chads or rich Saudi Oil Princes please rent GPU power to make feet loras for hunyuan I'm BEGGING
Anonymous 01/15/25(Wed)03:32:55 No.103901276
>>103901247
this, hunyuan can make perfect feet but it doesn't know the footjob or the foot licking concept
Anonymous 01/15/25(Wed)03:36:38 No.103901300
>>103901247
maybe if you prepare the dataset and labeled images
Anonymous 01/15/25(Wed)03:38:27 No.103901314
>>103900444
Indeed it's working... why is it the case? Did they train their model on looping gifs when they went to 201 frames?
Anonymous 01/15/25(Wed)03:47:04 No.103901356
>January 15th
>Still no hunyuan i2v
that doesn't look good at all...
Anonymous 01/15/25(Wed)03:53:14 No.103901395
>>103901356
if at some point we'll get the i2v model, will the t2v loras work on it or not?
Anonymous 01/15/25(Wed)04:18:34 No.103901547
1717883755629437
Anonymous 01/15/25(Wed)04:29:13 No.103901622
>>103901395
How the fuck should we know.
Anonymous 01/15/25(Wed)04:30:39 No.103901633
>9070xt
>still 16gb
Ohnononno ahahahaha
Anonymous 01/15/25(Wed)04:30:48 No.103901636
>>103901622
we already have local models that have both the t2v and the i2v models, and they have loras too, that's how the fuck we could know
Anonymous 01/15/25(Wed)04:49:24 No.103901757
>>103901633
wait, AMD never made a 24gb card?
Anonymous 01/15/25(Wed)05:00:37 No.103901816
>>103901757
You'll only have 16gb and you'll like it
Anonymous 01/15/25(Wed)05:01:26 No.103901819
>>103901816
this is insane, Nvdia is at 32gb already, it's like they want to loose lol
Anonymous 01/15/25(Wed)05:17:36 No.103901908
>>103901757
The 7900XTX but they didn't increase the bus width of the 9070 series die so it's stuck at 256 bit width. Clamshell it and you can do 32 GB but you will never get it for significantly below what Nvidia will charge you.
Somewhat on topic, despite the fact that we are using FP8 to FP4 now which has their own problems, I feel like if either AMD or Intel just releases a solid GPU for just inferencing with some training possible, at a low enough price, with the basic types FP32/FP16/BF16 and maybe TF32, people will swarm for it even if training support is poor. AMD unfortunately until very recently was missing BF16 support and Intel was MIA and had serious issues wih drivers and strange limitations on their first round of GPU designs.
Anonymous 01/15/25(Wed)05:19:56 No.103901922
imma bite the comfypill
what is the best browser to use it?
i remember brave had issues
Anonymous 01/15/25(Wed)05:29:24 No.103901988
>>103901922
there is no best browser because the frontend was completely ruined universally. it was uncomfy before, now it's uncomfy, laggy and unstable
Anonymous 01/15/25(Wed)05:34:53 No.103902024
>>103901988
wait what?
i thought comfy was the way to go now what happened?
so forge then?
Anonymous 01/15/25(Wed)05:39:03 No.103902060
>>103902024
It's always been cumfart. Forge doesn't work well with SD3.5 if you wanna try that and probably not that other one I forget the name, works with flux, sdxl, pony etc. I think it's dropped off in updates/support though.
Anonymous 01/15/25(Wed)05:42:10 No.103902082
>>103902060
>>103902024
Oh yeah, you have to modify one of the files in forge because it generates fake hashes for models and loras etc
Anonymous 01/15/25(Wed)05:46:42 No.103902126
>>103901922
>imma bite the comfypill
what do you want to use it for? if it's image models like sdxl or flux you could try one of the frontends that use the comfy api like swarmui or that stable diffusion plugin for krita
Anonymous 01/15/25(Wed)05:57:58 No.103902199
1711952413202819
https://xcancel.com/8co28/status/1879397900977099158#m
migu has boy hands here :d
Anonymous 01/15/25(Wed)06:00:32 No.103902215
1721777767981935
https://xcancel.com/toyxyz3/status/1879189674444624288#m
cosmos i2v seems all right, has anyone tested its nfsw capabilities on the 14b model?
Anonymous 01/15/25(Wed)06:06:36 No.103902255
>>103902199
Fizzledorf (the saviour of anime) btfo by a literally who
Anonymous 01/15/25(Wed)06:07:30 No.103902265
>>103902255
>Fizzledorf
who?
Anonymous 01/15/25(Wed)06:10:59 No.103902287
will you guys play PW's game
Anonymous 01/15/25(Wed)06:40:50 No.103902490
i am too cool for video games
Anonymous 01/15/25(Wed)07:18:21 No.103902759
space-elf-seated-cartoon
>>103901922
I use ungoogled chromium. Firefox gave me troubles with ComfyUI. But the new interface, though not without some nice features, is buggy and can be annoying. The mask editor especially is a pain.
Anonymous 01/15/25(Wed)07:22:59 No.103902792
>comfyUI
>uncomfortable af
Anonymous 01/15/25(Wed)07:23:27 No.103902794
>>103900720
correct comfy is a lying piece of trash, i also oom with like3 plenty of system ram still left over. and I'm pretty sure this shit has a memory leak because it will randomly for no reason start oom with settings that previously worked.
Anonymous 01/15/25(Wed)07:25:34 No.103902809
>>103901198
>illustrious
is fucking dog shit yeah...
Anonymous 01/15/25(Wed)07:26:29 No.103902813
>>103901922
Works fine for me in Vivaldi. Though sometimes it ooms while genning because I leave a couple hundreds of tabs open for weeks.
Anonymous 01/15/25(Wed)07:28:13 No.103902823
>>103901198
in fact its only for cartoon fags, i fucking hate those cartoon faggots, they belong in /sdg/, /ldg/ is for adults and not man babbies obsessing over a model that can do cartoon 2girls

fucking kys
Anonymous 01/15/25(Wed)07:29:14 No.103902833
Anonymous 01/15/25(Wed)07:32:23 No.103902850
>>103902823
noob would suit you well since it can generate a nice little pamper for you to wear
Anonymous 01/15/25(Wed)07:37:24 No.103902875
>>103901198
Base Illustrious looks good after upscaling with a fitting upscaling model and the second low denoise pass. Never had to use that "fixed" version.
Noob doesn't have that issue, the initial gens are smooth.
>perfect artstyle
Unless you use a lora you're locked to a single art style on Pony, everything looks samey. Illustrious can do a shit ton out of the box.
Anonymous 01/15/25(Wed)07:37:54 No.103902878
step_265000_validation_1024x1024
>lets see how the model is doing
1girl dataset and somehow it decides to march towards making a bloke
Anonymous 01/15/25(Wed)07:38:17 No.103902882
>>103902850
A fucking wot m8? you some kind of billy nonce?
Anonymous 01/15/25(Wed)07:43:06 No.103902925
>>103902882
im going to tickle your little billy if you keep crying on the adults only general, onga
Anonymous 01/15/25(Wed)08:06:02 No.103903126
>>103900353
How do you guys stay on top of new model releases?
Seems like there's so much stuff coming out constantly, I want to know when the next big thing comes out and I move on from Illustrious models.
Anonymous 01/15/25(Wed)08:20:47 No.103903218
>>103903126
>How do you guys stay on top of new model releases?
I go to /sdg/.
Anonymous 01/15/25(Wed)09:14:23 No.103903645
0
Anonymous 01/15/25(Wed)09:19:59 No.103903686
I'm giving up on comfy, its a buggy mess. it constantly fails to update nodes unless i update them manually, it also won't let me change their versons from nightly to latest. something i'm sure has a lot to do with recent changes to how the manager and the register works. the UI is bloat and laggy as fuck now.

NO NO NO NO, no more fucking UI bullshit, I'm gonna do everything on the command line.
Anonymous 01/15/25(Wed)09:21:55 No.103903701
>>103899948
Great job
Anonymous 01/15/25(Wed)09:22:54 No.103903707
>>103902215
alright if you consider her arm spasming ok
Cosmos has a major warble problem
Anonymous 01/15/25(Wed)09:31:02 No.103903783
00080-3122985308-d5682f4b-a0486e8bb6
Anonymous 01/15/25(Wed)09:31:44 No.103903796
0
Anonymous 01/15/25(Wed)09:35:46 No.103903824
I see Reddit is on the Flux cope phase. "Guys it's good, you just have to use Loras", "Guys, I swear it can do art, you just have to say everything except the art keywords related to that art style"
Anonymous 01/15/25(Wed)09:40:18 No.103903861
2 3090s or 1 5090?
Anonymous 01/15/25(Wed)09:44:04 No.103903889
>>103903861
I was reminded the other day how much slower my 3090 is compared to my 4090 when I was captioning with Lllava. One 5090 is going to be not only like 4x faster but 32 GB of VRAM will be a lot more versatile for training and inference.
Anonymous 01/15/25(Wed)09:49:26 No.103903943
>>103903218
where is sdg? vg? it's not in the op.
Anonymous 01/15/25(Wed)09:50:59 No.103903957
Anonymous 01/15/25(Wed)09:52:44 No.103903978
>>103902265
trani, trani btfo
Anonymous 01/15/25(Wed)09:59:21 No.103904031
0
Anonymous 01/15/25(Wed)10:11:54 No.103904146
0
Anonymous 01/15/25(Wed)10:12:07 No.103904148
>>103903126
90% of everything released is crap so there's no pressure.
Anonymous 01/15/25(Wed)10:16:06 No.103904180
>>103903889
What would 32GB actually unlock though? Hunyuan lora training?
Anonymous 01/15/25(Wed)10:41:50 No.103904429
Why wasn't my nogen in the collage?!??!?
Anonymous 01/15/25(Wed)10:42:38 No.103904433
HunyuanVideo_00020
I'm on a 12GB VRAm card
With 32GB of VRAM, can I gen Hunyan videos faster?
300 seconds is a long time to wait for a video
Anonymous 01/15/25(Wed)11:03:32 No.103904637
>>103903861
still can't split a diffusion model over two cards so
Anonymous 01/15/25(Wed)11:05:38 No.103904651
>>103903861
at that point why not an Ampere A6000 vs 5090
Anonymous 01/15/25(Wed)11:05:49 No.103904655
>>103904433
Sage Attention
fp8_fast
torch compile
Tea cache
Anonymous 01/15/25(Wed)11:06:56 No.103904666
1710143334878586
Anonymous 01/15/25(Wed)11:16:25 No.103904769
Long Dick General
Anonymous 01/15/25(Wed)11:17:06 No.103904777
>>103904651
One is $2k and the other is +$4.5k
Anonymous 01/15/25(Wed)11:24:18 No.103904856
>>103903824
They're only a couple months behind us, not bad.
Anonymous 01/15/25(Wed)11:55:08 No.103905210
1730656322289516
Anonymous 01/15/25(Wed)11:56:29 No.103905232
>>103902809
yeah noob is much better
Anonymous 01/15/25(Wed)11:57:20 No.103905248
Video
>>103900353
For some reason ComfyUI keeps crashing after the video sampling is done, here is my setup, i copied from the mario one in the rentry. Anyone has any ideas of what i may be doing wrong? I even tried scaling it down slightly to make sure that running out of VRAM wasn't the issue
Anonymous 01/15/25(Wed)11:58:10 No.103905263
How much of a disadvantage do RX 6000 GPUs have against RTX 3000 GPUs in performance?
Anonymous 01/15/25(Wed)11:58:15 No.103905265
>>103902823
show us ur cool gens, anon!
Anonymous 01/15/25(Wed)12:02:30 No.103905314
>>103905210
Sick gen
Anonymous 01/15/25(Wed)12:24:38 No.103905563
>>103901819
>it's like they want to loose lol
It's almost as if they are controlled opposition. No... that can't be...
Anonymous 01/15/25(Wed)12:38:24 No.103905728
hibernation mode
Anonymous 01/15/25(Wed)12:43:14 No.103905798
There's been strides in online resources to make AI videos it seems but a lot of it is sfw or paid.
Is it ready yet to be able to do anything for the free models like A1111 or not at all?
Last time, the only viable option was either rotoscopy or same frame loras. And this required upwards of 30 gb vram. Anything viable for a 4070Ti super-let?
Anonymous 01/15/25(Wed)12:55:00 No.103905963
>>103905248
Dunno, my only suggestion would be find a newer workflow that mario gen is quite old at this point and there's been lots of updates to hunyuan wrapper since then. Alternatively you could try comfy's nodes
Anonymous 01/15/25(Wed)12:59:55 No.103906049
Fuck that Razzz guy spamming his shitty overtrained celebslop loras. Every time I check the latest there's 10 new ones
Anonymous 01/15/25(Wed)13:00:48 No.103906061
Untitled
>>103905248
if this radio switch is checked then it'll ignore the values you put in above it and use the defaults
Anonymous 01/15/25(Wed)13:07:33 No.103906150
>>103906049
Kino taste in women tho
Anonymous 01/15/25(Wed)13:37:05 No.103906548
>>103902878
Which model? Your own?
Anonymous 01/15/25(Wed)14:10:40 No.103906967
not to be crude but what's the correct terminology for a strap on
Anonymous 01/15/25(Wed)14:10:55 No.103906973
00024-1947643488
Anonymous 01/15/25(Wed)14:13:18 No.103907009
>>103906967
run it through joycap and see
Anonymous 01/15/25(Wed)14:14:31 No.103907031
ComfyUI_10109_
Anonymous 01/15/25(Wed)14:19:55 No.103907098
>>103907009
good call
>She wears a phallic sex toy attached to a harness, which she grabs in her hand
and of course hunyuan already knows what it is
Anonymous 01/15/25(Wed)14:20:53 No.103907117
S1448007797; ntrMIXIllustriousXL_xiii; DPM++ 2M SDE; 20250114125730;
i really hate ears
Anonymous 01/15/25(Wed)14:33:42 No.103907275
https://civitai.com/models/1144849/anime-tentacle-sex-hunyuan?modelVersionId=1287585
that one looks really high quality, he probably trained on a very high resolution with a lot of frames
Anonymous 01/15/25(Wed)14:36:32 No.103907306
>>103906049
I blocked that motherfucker to not see him anymore lol
Anonymous 01/15/25(Wed)14:37:34 No.103907321
1735478699206841
https://civitai.com/models/1135830/facial-cumshot?modelVersionId=1277065
this is ridiculous, now there's loras that only works on kijai and not comfy core? COMFYUI IS A FUCKING MEEEEEEEEMEEEEEEEEEE
Anonymous 01/15/25(Wed)14:41:05 No.103907364
Capture
>>103907275
>on a very high resolution
Nope, 244x244. If it was trained with musubi you can check the meta data
>>103907321
You just have to run picrel after training with musubi. I made that mistake on my first couple loras
Anonymous 01/15/25(Wed)14:42:58 No.103907389
>>103907364
>You just have to run picrel after training with musubi.
oh ok thanks, I wonder why it's not automatic though, like why should you have to do this manually, of course this shit has to be compatible with comfy core
Anonymous 01/15/25(Wed)14:45:01 No.103907418
1733370942055275
>Pictured: The last Human confronting the Singularity that destroyed humanity
Anonymous 01/15/25(Wed)15:02:14 No.103907614
>>103907364
Does this affect the quality of the LoRA?
Anonymous 01/15/25(Wed)15:15:30 No.103907784
>>103907614
not at all, it's just a conversion so that comfy core recognize it, it doesn't change anything else
Anonymous 01/15/25(Wed)15:18:15 No.103907819
>>103907418
>oi... give over.
Anonymous 01/15/25(Wed)15:24:29 No.103907894
00010-1155281705
Anonymous 01/15/25(Wed)15:26:42 No.103907924
>>103907894
Did you prompt for "polished ass" sir?
Anonymous 01/15/25(Wed)15:37:32 No.103908055
Moar sfumato.
Anonymous 01/15/25(Wed)15:39:52 No.103908090
1706922739081830
https://lumalabs.ai/ray
the Video AI ecosystem is so competitive, Sora looks like mid tier now with all those new models
Anonymous 01/15/25(Wed)15:42:18 No.103908136
Is it possible to run flux or similar on a laptop with a 1660 Ti 6GB? I tried running it just now, but RAM (not VRAM) capped out and I got OOM'd with my 16 GB. I could upgrade to 64 GB if it would help, but I don't feel like wasting money if it still doesn't work.
Anonymous 01/15/25(Wed)15:43:28 No.103908146
>>103908136
>6GB
nah won't be enough, buy a better gpu
Anonymous 01/15/25(Wed)15:44:38 No.103908166
>>103908136
Buy one of these : https://www.nvidia.com/en-eu/project-digits/

Ask your boss for a bonus and get it.
Anonymous 01/15/25(Wed)15:44:56 No.103908171
>>103908146
Alright. I'll move the setup to my gaming computer instead.
Anonymous 01/15/25(Wed)15:46:00 No.103908185
>>103908166
>inference large AI models of up to 200B parameters locally
oof, temping, but still not good enough to run minimax
Anonymous 01/15/25(Wed)15:46:51 No.103908198
>>103908136
>I tried running it just now, but RAM (not VRAM) capped out
Make your swapfile larger
Anonymous 01/15/25(Wed)15:47:48 No.103908209
What's a good model for someone who's tired of making 1girl stuff, but too poor to run flux? I'm using RealVisXL right now, and while the outputs look nice it doesn't seem as flexible as I'd like.
Anonymous 01/15/25(Wed)15:48:44 No.103908223
Anonymous 01/15/25(Wed)15:48:59 No.103908226
>>103908198
>Make your swapfile larger
you want this anon to kill his SSD, just ask him to buy some more ram it's kinda cheap
Anonymous 01/15/25(Wed)15:49:02 No.103908227
>>103908209
https://civitai.com/models/201287/iris-lux-polyvalent-prototyperealisticsfwartnsfwpornno-refiner-needed
Anonymous 01/15/25(Wed)15:51:34 No.103908258
00015-2163732038
>>103907924
she was born with it!
Anonymous 01/15/25(Wed)15:55:18 No.103908292
Anonymous 01/15/25(Wed)16:00:17 No.103908363
file
>this is what these NIGGERS want me to use instead of forge
ROLF XDDDDDD LMAO
THE FUCK IS THIS????
DID A 4 YEAR OLD DESIGN THIS PAGE?????
HAHAAHAHAHAHAHAHAHAHAH XD
Anonymous 01/15/25(Wed)16:02:04 No.103908381
1715047517508074
>>103908363
>he thinks the gradio backdoor app is an example of a good Ui
Anonymous 01/15/25(Wed)16:02:33 No.103908387
>>103908363
Looks okay.
Anonymous 01/15/25(Wed)16:04:59 No.103908420
00029-1935747534
Anonymous 01/15/25(Wed)16:10:09 No.103908497
>>103908363
if you don't like the look you can go to the "comfyui workflow" tab like a real man
Anonymous 01/15/25(Wed)16:11:43 No.103908518
>>103908497
The SpaghettiOs of UI.
Anonymous 01/15/25(Wed)16:12:51 No.103908529
1720080727862108
>>103908363
what's wrong with SwamUi? It's still better than having to deal with this spaggheti shit
Anonymous 01/15/25(Wed)16:13:14 No.103908533
>>103908518
imagine having to code that spaghetti yourself instead of just wiring up existing nodes.
Anonymous 01/15/25(Wed)16:13:56 No.103908541
ComfyUI_10298_
Anonymous 01/15/25(Wed)16:17:25 No.103908584
>>103908518
spaghettiOs are unironically good tho
Anonymous 01/15/25(Wed)16:26:43 No.103908704
trouble jewing
>>103902875
I feel like the opposite is true here. Illustrious gens look high quality but have a overly generic anime theme that is perfect for avatarfags from /sdg/.

You can't do shit other than have your generic waifu stand there and smile compared to Pony which has a lot of sex/poses based on booru tagging built in.
Anonymous 01/15/25(Wed)16:33:41 No.103908768
>>103906049
I'm surprised Civitai doesn't require buzz to post more than one model in a given 24 hour basis as posters can just spam low effort or badly trained models with lower epochs.

The ratings system also doesn't offer any good way to filter out trash since negative ratings are non visually present.
Anonymous 01/15/25(Wed)16:37:56 No.103908819
1722604729429517
all of this just for this?
https://civitai.com/images/51848069
Anonymous 01/15/25(Wed)16:38:40 No.103908827
>>103908584
>soggy noodles in a shitty canned tomato soup are unironically good
terrible b8
Anonymous 01/15/25(Wed)16:40:55 No.103908852
How to fix Civitai
Get rid of the whole buzz eco system
No more on site training
Ban celebrity loras
IP block India
Anonymous 01/15/25(Wed)16:41:41 No.103908865
>>103908704
this is why hunyuan i2v must be a thing, it's just more fun
Anonymous 01/15/25(Wed)16:43:16 No.103908889
>>103908852
>Ban celebrity loras
why? you're a swiftie?
Anonymous 01/15/25(Wed)16:43:38 No.103908895
>>103908852
>how to fix the internet
>ip block india, australia and new zealand
ftfy
Anonymous 01/15/25(Wed)16:45:01 No.103908913
>>103908865
>this is why hunyuan i2v must be a thing
https://github.com/Tencent/HunyuanVideo/issues/180#issuecomment-2592674405
Finally we got some feedback from them
>We are working hard on developing a text-to-video model and believe it will be completed soon. However, the open-sourcing process involves some steps that are beyond our control, so we cannot estimate the exact timeline. We believe that it will be available to everyone soon.
that doesn't sound too good...
Anonymous 01/15/25(Wed)16:46:40 No.103908935
>>103908913
>a text-to-video model
uh...
Anonymous 01/15/25(Wed)16:46:49 No.103908940
>>103908913
>a text-to-video model
what? the issue was about the i2v model? is that a typo?
Anonymous 01/15/25(Wed)16:48:36 No.103908964
>>103908889
No but having all the resources easily available for any jeet or retard to start pumping out ai celebslop deepfake porn is how this shit gets shut down
Anonymous 01/15/25(Wed)16:49:49 No.103908976
>>103908964
it's been 2 years that celebrities loras are a thing on civitai, and this site is still doing fine, living like a frightened coward is never the way to go anon
Anonymous 01/15/25(Wed)16:50:26 No.103908982
>>103908913
would you give it a rest, they said january and we're only half way through
I get you're excited but have some darn patience
Anonymous 01/15/25(Wed)16:51:35 No.103908995
>>103908982
>would you give it a rest
nope, Mochi betrayed us, so I can't be optimistic for hunyuan now
Anonymous 01/15/25(Wed)16:51:46 No.103908996
HunyuanVideo_00431
>>103908865
oh its fun trying to make it work currently with spaghetti ass workflows. people are hoping for an official support release but honestly it'll probably get shelved in terms of support as no company will ever actually open-source god-tier i2v models.

>>103908895
wtf did the kiwis do?
Anonymous 01/15/25(Wed)16:52:40 No.103909007
>>103908976
Yes but we didn't have a completely uncensored video model for most of those two years.
Anonymous 01/15/25(Wed)16:53:01 No.103909010
>>103908996
>no company will ever actually open-source god-tier i2v models
People have been saying this about all kinds of AI models for years now and local is still chugging along.
Anonymous 01/15/25(Wed)16:53:42 No.103909018
>>103909007
we had the same discussion 2 years ago
>oh no we can make realistic deepfake porn images now, it's the end of the world, civitai is gone in 2 weeks, Taylor Swift will burn it to the ground
Anonymous 01/15/25(Wed)16:54:55 No.103909036
>>103909018
Flux can't do porn and I have never seen a convincing SDXL realistic porn image
Anonymous 01/15/25(Wed)16:55:08 No.103909039
>>103908996
>no company will ever actually open-source god-tier i2v models
so what's the prediction for you? Tencent won't release hunyuan i2v? or if they release it it'll be a cucked piece of shit?
Anonymous 01/15/25(Wed)16:56:09 No.103909056
>>103909036
oh boy you definitely need to go on /b/ more often anon
>>>/b/930041555
Anonymous 01/15/25(Wed)16:57:42 No.103909075
>>103909056
You're fucking blind if you think any of that garbage would fool anyone but the dumbest of jeets
Anonymous 01/15/25(Wed)16:58:19 No.103909079
>>103909075
then what's the problem? hyvid ain't realistic either, not even close
Anonymous 01/15/25(Wed)16:58:56 No.103909085
>>103909075
>he didn't hear about the retarded boomer who sent a jeet $800,000 using poorly photoshopped brad pitt images
Anonymous 01/15/25(Wed)16:59:10 No.103909090
>>103908819
kek, just follow kijai workflow or comfy, that autistic spaghetti mess is just bad
Anonymous 01/15/25(Wed)16:59:35 No.103909100
1713444909956252
>>103909075
dude, some women wasted hundreds of thousands of dollars because of this SD1.4 level kind of picture, you realized just now that people are gullible retards?
https://xcancel.com/nexta_tv/status/1879171365678289217
Anonymous 01/15/25(Wed)17:00:38 No.103909110
>>103909075
>anyone but the dumbest of jeets
so 95% of the population? that's a good number if you ask me
Anonymous 01/15/25(Wed)17:03:15 No.103909153
file
ok decided to bite the bullet!
learning swarmAI with this fellow's videos!
Anonymous 01/15/25(Wed)17:04:14 No.103909163
>>103909153
learning something from the turkish god is always a privilege, have some fun with this fine gentleman anon
Anonymous 01/15/25(Wed)17:05:59 No.103909181
>>103909079
>hyvid ain't realistic either, not even close
I disagree, >>103908819 this looks far more realistic than anything you'll get from SDXL
Anonymous 01/15/25(Wed)17:06:24 No.103909185
>>103909039
they will likely release a 'lite' version to keep their word about being a defacto champion for open source but will likely branch off with a refined model that will be another pay2gen site like all the other closed source i2v models out there.They will also just keep delaying publishing their findings to keep riding the AI money train (VC / investors) for as long as they can without needing to deliver a ' final' product.

It makes no sense for a such a giant profit driven conglomerate like Tencent to release decent shit for free other than to maybe gain market share or convert local diffusionbros into potential paypigs that aren't subscribed to any existing site.
Anonymous 01/15/25(Wed)17:07:30 No.103909198
1735673827096704
>>103909181
I think flux is still the model that can make the most realistic outputs of all models
https://civitai.com/models/652699/amateur-photography-flux-dev
Anonymous 01/15/25(Wed)17:07:52 No.103909201
>>103909181
Then you don't necessarily disagree. More realistic than SDXL does not equal realistic, it's not a grading system, it's a hard binary
Anonymous 01/15/25(Wed)17:08:30 No.103909208
>>103909185
>they will likely release a 'lite' version to keep their word about being a defacto champion for open source
so like the t2v model we're actually using? because we only got the distilled model, they kept the real base model on their API, it's flux all over again
Anonymous 01/15/25(Wed)17:10:07 No.103909222
>>103909036
>and I have never seen a convincing SDXL realistic porn image
You must be very new
Anonymous 01/15/25(Wed)17:12:02 No.103909240
ComfyUI_10347_
Anonymous 01/15/25(Wed)17:14:40 No.103909266
>>103909208
the age of hobbyist finetunes is over, even with the original weights out in the open, nobody but tencent themselves can do anything with them
Anonymous 01/15/25(Wed)17:16:05 No.103909281
>>103909198
Even with that lora I can still see the Flux face. You can never not notice it once you're familiar with it
Anonymous 01/15/25(Wed)17:17:04 No.103909290
>>103909281
the thing is that the vast majority of people who doomscrool on facebook will never be familiar with the flux face so to them they just look like regular girls
Anonymous 01/15/25(Wed)17:20:04 No.103909325
1720184939027806
>>103908913
>We are working hard on developing a text-to-video model
Anonymous 01/15/25(Wed)17:20:21 No.103909327
>>103909198
>FTNA
Why are they doing open bobs?
Anonymous 01/15/25(Wed)17:21:18 No.103909337
>>103909208
everything is going to be smoke and mirrors for any AI development since realistically you're fucked as soon as you show your all your cards.

they'll probably release a decent t2v model but i2v will be cucked since there is liability with ownership of the original image used. all of the genning sites have pretty strict filters and are ready to shit the bed if any government or celeb like Taylor Swift goes after them in full force.

there's already a lot of legislation in the works since high schoolers keep genning lewds of girls in their classes
Anonymous 01/15/25(Wed)17:22:36 No.103909348
>>103909337
>they'll probably release a decent t2v model but i2v will be cucked
with that logic the t2v model should've been cucked aswell because we can do some really realistic celebrities loras out of it, we can also do really good porn, mix the both of them and then...
Anonymous 01/15/25(Wed)17:23:11 No.103909356
>>103909337
Don't fud my coin bro
Anonymous 01/15/25(Wed)17:23:56 No.103909366
>>103909337
t2v is already v2v capable, that box is already open
Anonymous 01/15/25(Wed)17:24:05 No.103909367
>>103909337
>they'll probably release a decent t2v model
it's already there anon, and it's fully uncensored, and we can do celebrities with it, the genie is out
Anonymous 01/15/25(Wed)17:25:20 No.103909384
>>103909337
>there's already a lot of legislation in the works since high schoolers keep genning lewds of girls in their classes
AI is being used as a scape goat here. Those nudify web apps have existed for years high schoolers aren't loading up comfyui on their phones and proompting while in class
Anonymous 01/15/25(Wed)17:26:17 No.103909396
>>103909266
>the age of hobbyist finetunes is over
yep, the BFL business plan is perfect, they have the most powerful model in their server, we only got an inferior version, and because it's distilled we can't finetune and improve it so we can't beat the API, we have no choice but to pay for the API if we want superior outputs, evil genius
Anonymous 01/15/25(Wed)17:27:43 No.103909412
>>103909396
or you'd be paying a server farm to finetune the undistilled model
that's what I'm saying, it's all too expensive to be practical
Anonymous 01/15/25(Wed)17:28:00 No.103909419
>>103909337
>since there's bad people in this world we should forbid every piece of technology ever to people
good luck trying to remove the computer, the internet, the camera, photoshop... to us, why are you pretending that only AI is capable to do some nasty shit like that?
Anonymous 01/15/25(Wed)17:29:02 No.103909429
>>103909412
>it's all too expensive to be practical
dunno about that, the llm fags have to deal with much bigger models than us and they're still finetuning their giant models
Anonymous 01/15/25(Wed)17:44:51 No.103909567
>follow OP guide on how to install on Windows with AMD card
>warning/error/warning/error
so is there a guide that works is it just eaiser to make a Linux virtual machine
Anonymous 01/15/25(Wed)17:45:21 No.103909569
>>103909419
not saying those are the only things AI is capable of. Just saying the general spiel of what normies/politicians that don't get on the internet much get told about AI. Porn is also a great motivator for both development and detterence. Would there be any actual interest in Hunyuan if it was NOT uncensored?

while of course we will always have the tools to continue working with AI there's lot of opposing forces out there that want you to own nothing and pay a subscription as well as never be able to understand / reverse engineer the toolsets as a hobbyist.
Anonymous 01/15/25(Wed)17:46:27 No.103909577
>>103909567
>with AMD card
anonie....
Anonymous 01/15/25(Wed)17:46:31 No.103909578
>>103909569
>Would there be any actual interest in Hunyuan if it was NOT uncensored?
of course, Flux is censored and this model is still the most hyped model ever at its release
Anonymous 01/15/25(Wed)17:47:54 No.103909590
>>103909567
switch to loonix and use rocm trust me on this one anon
Anonymous 01/15/25(Wed)17:48:27 No.103909596
>>103909567
>Windows with AMD card
buy a fucking Nvdia card anon...
Anonymous 01/15/25(Wed)17:49:25 No.103909604
>>103909569
>Would there be any actual interest in Hunyuan if it was NOT uncensored?
not all of us use it exclusively for porn, it's a very capable model and it knows more things than flux out of the box, like hair curlers for example, a pretty common thing but flux just cannot do it, neither can SD actually. That's what I value most, versatility, and if that gets lost when censoring boobies then keep the damn boobies
Anonymous 01/15/25(Wed)17:54:06 No.103909637
>>103909577
>>103909590
>>103909596
why did I spend 700 dollars on this....
Anonymous 01/15/25(Wed)17:54:51 No.103909647
HunyuanVideo_00018
>>103909604
This. Flux complete failure to understand human anatomy when it's not a basic standing or sitting pose kills the model for me
Anonymous 01/15/25(Wed)17:55:39 No.103909661
>>103909637
if it was important you would have looked into it
Anonymous 01/15/25(Wed)17:56:04 No.103909666
>>103909647
Can you regen this but using the BBW LoRA that's on civit?
semi-nsfw: https://civitai.com/models/1144518/ga1n3rb0t?modelVersionId=1287232
Anonymous 01/15/25(Wed)17:56:16 No.103909668
>>103909637
windowsfags need to suffer
Anonymous 01/15/25(Wed)17:57:08 No.103909677
>>103909647
yeah, hunyuan is insane at anatomy, thanks to nudity and porn, the chinks aren't stupid they know that giving the model complex poses through porn would help the model
Anonymous 01/15/25(Wed)17:57:11 No.103909680
Not using Linux for this is like shooting yourself in the foot and then shooting yourself in the head
Anonymous 01/15/25(Wed)17:58:15 No.103909699
>>103909661
I saw AMD + Windows in the comfy ui install thing and thought that meant I could get Windows and an AMD card
Anonymous 01/15/25(Wed)17:58:29 No.103909701
1732657929875755
>>103909680
nah you're full of shit, you can do everything on windows that you can on linux, fuck your troonix shit
Anonymous 01/15/25(Wed)17:58:52 No.103909706
>>103909699
but you didn't read it?
Anonymous 01/15/25(Wed)17:59:16 No.103909711
i pray for windows anon
Anonymous 01/15/25(Wed)17:59:23 No.103909714
>>103909680
even jacket man said linux is best AAIIIEEEEE
Anonymous 01/15/25(Wed)18:00:12 No.103909723
a1111 outputs
what your output folder stats?
Anonymous 01/15/25(Wed)18:02:18 No.103909742
windowsfren
>>103909637
tale as old as time windowsfren, having AMD is basically choosing hard mode for everything AI-related unfortunately since their software support is still super bad.

while everyone is saying 'use loonix' it may be salvageable on windows if you have time to listen to indians on youtube for hours to help you navigate initial configs.
Anonymous 01/15/25(Wed)18:02:24 No.103909743
>>103909701
>you can do everything on windows that you can on linux,
but can you customize windows like you can customize your body is the question (and the answer is no)
Anonymous 01/15/25(Wed)18:03:35 No.103909748
>>103909743
why should I give a fuck? the default Ui on windows works fine, it's more convenient to spend your time on windows than on troonix
Anonymous 01/15/25(Wed)18:04:29 No.103909763
>>103909748
>inserts advertisements into your start menu
heh, nothing personal kiddo
Anonymous 01/15/25(Wed)18:04:58 No.103909770
>>103909743
>like you can customize your body
something a troon would say, not surprising from a troonix user
>>103909763
>>inserts advertisements into your start menu
they can be removed, and I removed them, see, here's your customization
Anonymous 01/15/25(Wed)18:05:31 No.103909778
firefox_hpW6rGb9Ew
>>103909706
I don't understand what part of the guide here says 'dont try window+AMD', if anything it says the opposite

>>103909742
I will try but holy moly is this frustrating
Anonymous 01/15/25(Wed)18:06:18 No.103909787
>>103909778
first line dude
Anonymous 01/15/25(Wed)18:07:25 No.103909800
>>103909770
but anon loonix doesnt even have ads on the "start menu" to begin with
next youll start complaining about being filtered by venvs....
Anonymous 01/15/25(Wed)18:07:37 No.103909802
>>103909743
why does this make the trans-linux correlation make so much more sense when you put it that way.
Anonymous 01/15/25(Wed)18:08:58 No.103909820
1716039359769159
Anonymous 01/15/25(Wed)18:09:15 No.103909824
>>103909778
anon
you will suffer with windows + amd
dual boot something like ubuntu and use the rocm variant
im not trolling fren
Anonymous 01/15/25(Wed)18:09:19 No.103909826
>>103909800
>but anon loonix doesnt even have ads on the "start menu" to begin with
but I thought you were happy about customization, if you think so highly about customization it means that the default troonix settings weren't at your taste and you wanted to change some stuff, how is that different to me who wanted to customize windows because its default settings weren't at my taste either? your estrogen bias is showing anon
Anonymous 01/15/25(Wed)18:10:41 No.103909837
>>103909826
>windows customization is at the same level as loonix
anon i...
Anonymous 01/15/25(Wed)18:13:01 No.103909871
>>103909837
you're moving the goalpost, your point was that I had to customize windows to enjoy it
>>103909800
>but anon loonix doesnt even have ads on the "start menu" to begin with
you made it seem it's a bad thing even though you just said 5 mn earlier that the appeal of troonix is the fact you can customize stuff >>103909743
>but can you customize windows like you can customize your body
disingenuous troon
Anonymous 01/15/25(Wed)18:13:06 No.103909873
1720454431059294
Anonymous 01/15/25(Wed)18:15:06 No.103909898
>>103909871
dont take your frustrations out on me anon i know its annoying to have to run some third party spyware in order to remove windows bloat thats why i recommend loonix
Anonymous 01/15/25(Wed)18:16:43 No.103909914
>>103909898
>i know its annoying to have
I thought customizing your operating software was cool and all, so when you customize troonix it's cool and le heckin wholesome, but when you customize windows it's annoying? your estrogen pills are frying your brain anon
Anonymous 01/15/25(Wed)18:19:25 No.103909952
Anonymous 01/15/25(Wed)18:22:47 No.103909993
https://civitai.com/models/1131397?modelVersionId=1280811
>so many cope """"i2v"""" workflows
is the i2v hunyuan model the most anticipated model of all time?
Anonymous 01/15/25(Wed)18:24:01 No.103910005
>>103909993
>V3!!! XL Image 2 Video HunYuan Janky Workflow: Best until the real deal
At least he's sort of honest about it not being actual i2v unlike the others kek
Anonymous 01/15/25(Wed)18:25:33 No.103910013
https://files.catbox.moe/vvbglx.mp4
(sfw but not polite company)
I fucking love this model so much
try to make this in vanilla flux, or find it on shutterstock, I double dare you
Anonymous 01/15/25(Wed)18:26:21 No.103910018
1722706076266117
Time travelers and their cool macbooks.
Anonymous 01/15/25(Wed)18:26:36 No.103910019
>>103910013
>try to make this in vanilla flux
you can't do an angry standing grandma with this pink plastic thing on flux?
Anonymous 01/15/25(Wed)18:27:24 No.103910026
>>103909723
You don't need all that and you know.
Anonymous 01/15/25(Wed)18:28:27 No.103910032
>>103909993
>is the i2v hunyuan model the most anticipated model of all time?
Nah, that would be SD3.
Anonymous 01/15/25(Wed)18:29:36 No.103910042
>>103910032
the thing with SD3 is that it took forever to be released, so it was intense because it was a long wait, I hope we won't have to wait that long for hunyuan i2v
Anonymous 01/15/25(Wed)18:30:30 No.103910047
1720613580175678
>>103909952
Literally what you see on the picture.
Anonymous 01/15/25(Wed)18:31:06 No.103910053
>>103910042
Makes it all the funnier how shit it was and how fast it was dropped
Anonymous 01/15/25(Wed)18:31:49 No.103910062
unironically two weeks until chinese new year
Anonymous 01/15/25(Wed)18:32:32 No.103910068
file
>first gen with swarm
uh???
the fuck this shit?!
Anonymous 01/15/25(Wed)18:32:43 No.103910070
>>103910053
I think it was even more brutal to them during the SD3.5 release, you can feel that on SD3 they haven't tried hard to make a good model, they were a monopoly so they thought they could release a turd and we would eat it up, but after the huge backlash and the Flux release, they decided to try harder and make something acceptable, and then we saw SD3.5 and we dropped it like in less than a week kek
Anonymous 01/15/25(Wed)18:32:54 No.103910072
>>103910062
lunar calendar freaks, how dare they have their own customs
Anonymous 01/15/25(Wed)18:33:16 No.103910079
>>103910047
Fine. Keep your secrets. But please continue to post these kino outputs.
Anonymous 01/15/25(Wed)18:34:35 No.103910094
>>103909778
There are two main issues with AMD + Windows for AI.
1.) ROCm releases that support Windows at https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html which is at 6.2.4 is behind where Linux is which according to https://rocm.docs.amd.com/en/latest/ is 6.3.1
2.) What you get with the HIP SDK is a subset of ROCm that AMD managed to port to Windows. So you are going to run into edge cases.
Anonymous 01/15/25(Wed)18:35:07 No.103910100
>>103910062
That's it. It either happens in the next week +3 days or we're not getting img2vid in January. They can't release hyvid img2vid right before the new year or there will be nobody to put out fires.
Anonymous 01/15/25(Wed)18:35:21 No.103910103
DOES SWARMAI HAVE A GELBOORU PROMPT AUTOCOMPLETE?
else i finna really have to stay at forge
Anonymous 01/15/25(Wed)18:36:22 No.103910114
>>103910103
it does but youre not gunna like how to get it
its easy with the comfyui manager but you have to install it first and considering youre using swarm i imagine you dont like to mess around with things very much
github search comfyui manager
Anonymous 01/15/25(Wed)18:37:56 No.103910124
If people spent 1/100th of the time they spent avoiding comfy UI and complaining about it they'd have basic mastery over the UI already.
Anonymous 01/15/25(Wed)18:39:13 No.103910138
>>103910114
i actually used comfyui once before but i prefer to have a interaface, at least for now i am too much of a newbie.

btw i dunno if it's placebo but why the fuck do images literally look better on swarm than they did on forge?
using the same model even wtf this witchery
Anonymous 01/15/25(Wed)18:39:29 No.103910141
1736009124008001
>>103910079
No, there's literally no secrets
It's just
>very awa, masterpiece, media type, an artist or two at random weights, 1something, massive tits, hairy anus, fucked by a giant horse cock, an artist technique or a shot type
Noob is doing all the heavy lifting
Seriously i'm amazed about the capabilities of this model
Anonymous 01/15/25(Wed)18:41:44 No.103910152
>>103910138
tokens are weighted differently (you can install a custom node which emulates auto version tho), and i think auto uses cpu noise while comfy uses gpu
everything outside of that is placebo
you have now used twice your allotted newfag questions per thread
Anonymous 01/15/25(Wed)18:43:50 No.103910173
>>103910152
ok
do you think i should use swarm or comfy?
i hear comfy is trash on browser...
last time i used i even had a issue where some of the buttoms looked kinda stretched on chrome as if they were low res
Anonymous 01/15/25(Wed)18:46:38 No.103910193
>>103910103
You just need to install it. Yes, it does have it.
Anonymous 01/15/25(Wed)18:50:31 No.103910225
1722840477330222
Anonymous 01/15/25(Wed)18:51:45 No.103910238
>>103910141
>Seriously i'm amazed about the capabilities of this model
I can't see another local anime beating it unless it has a 16ch VAE desu
Anonymous 01/15/25(Wed)18:52:22 No.103910242
>>103910238
>I can't see another local anime beating it unless it has a 16ch VAE desu
like NAI v4?
Anonymous 01/15/25(Wed)18:54:31 No.103910263
>>103910242
I don't know what happened with v4 but Anon definitely wasn't wowed by its outputs despite its higher channel encoder.
Anonymous 01/15/25(Wed)18:55:34 No.103910273
1000103922
>>103900353
Whats a good checkpoint to begin with? Using comfy-zluda if that matters
Anonymous 01/15/25(Wed)18:56:46 No.103910288
Noob question.

Can I take let's say some kino hentai picture and change the character into someone else in Forge?
Anonymous 01/15/25(Wed)18:57:30 No.103910294
1729822077337544
>>103910238
What's pony author is even doing right now? This shit is like next gen compared to pony.
Anonymous 01/15/25(Wed)18:59:50 No.103910314
>>103910294
>What's pony author is even doing right now?
removing more artist names so that it's even safer than before!
Anonymous 01/15/25(Wed)19:01:42 No.103910325
>>103910314
I'd wager he's given up or he's put his time into something worthless.
Anonymous 01/15/25(Wed)19:05:46 No.103910355
>>103910294
https://civitai.com/articles/6309/towards-pony-diffusion-v7-going-with-the-flow
Back when he did that live stream with Civitai the thread was lambasting it. But the short version is he's retarded.
Anonymous 01/15/25(Wed)19:10:54 No.103910404
ComfyUI_00004_
Anonymous 01/15/25(Wed)19:12:23 No.103910419
1717844230445903
what the fuck have i prompted
Anonymous 01/15/25(Wed)19:14:53 No.103910449
1716792493227312
Anonymous 01/15/25(Wed)19:18:36 No.103910490
>>103910273
lurk moar
Anonymous 01/15/25(Wed)19:25:17 No.103910562
1735009942091691
>>103910173
>do you think i should use swarm or comfy?
Swarm has built in comfy because it's literally a comfy based web interface
Its massive benefit is having a humane interface with a ton of quality of life functions like ability to pull lora info directly from civitai
Anonymous 01/15/25(Wed)19:41:33 No.103910765
Still waiting on my BBW doing a handstand
=[
Anonymous 01/15/25(Wed)19:42:52 No.103910775
>>103910765
Big Bad Weetleborgs?
Anonymous 01/15/25(Wed)19:45:54 No.103910815
>>103910775
Disgusting fat women bro, it would be funny
Anonymous 01/15/25(Wed)19:46:16 No.103910821
00168-0
Hello everyone, this is my character, Red Witch. I intend to post her over and over and over every single day without even bothering to consider whether or not people might be tired of seeing her. My prompt will always be "1girl just standing there not doing anything interesting that would make the image into a noteworthy picture worth sharing on four channel dot org."
Anonymous 01/15/25(Wed)19:46:25 No.103910823
Hunyuantest
is it possible to fix this weird thing?
Anonymous 01/15/25(Wed)19:46:53 No.103910830
>>103910823
hyvid best model no flaws western dog
Anonymous 01/15/25(Wed)19:47:18 No.103910835
>>103900502
unholy faketits
Anonymous 01/15/25(Wed)19:48:15 No.103910849
>>103904666
I normally just shitpost, but I want to say that this is a really fucking cool image, dude
I kinda wish nobody would post unless they had something this dope to share, because then these threads would always be 10/10
Anonymous 01/15/25(Wed)19:48:22 No.103910852
>>103910830
lmao, I read it with some chink voice in my head it was funni
Anonymous 01/15/25(Wed)19:49:23 No.103910864
>>103910849
>I kinda wish nobody would post unless they had something this dope to share, because then these threads would always be 10/10
this
Anonymous 01/15/25(Wed)19:49:23 No.103910865
>>103910849
>I kinda wish nobody would post unless they had something this dope to share
We can only lead by example, which I do. NTA tho
Anonymous 01/15/25(Wed)19:53:17 No.103910904
0
Anonymous 01/15/25(Wed)19:54:11 No.103910914
Are you, too, addicted to slopping?
Anonymous 01/15/25(Wed)19:54:40 No.103910920
Addicted to that kinosovl. Straight to the veins.
Anonymous 01/15/25(Wed)19:55:16 No.103910927
Anonymous 01/15/25(Wed)19:55:24 No.103910928
ComfyUI_00015_
>>103910821
Do you also intend to roleplay with equally annoying avatarfags, destroying threads with bullshit in the process?
Anonymous 01/15/25(Wed)19:56:34 No.103910939
>>103910823
You're using the res_multistep sampler right? More steps would help too.
Anonymous 01/15/25(Wed)19:57:34 No.103910950
Anonymous 01/15/25(Wed)19:57:45 No.103910954
>>103910821
Can I make my avatar (Yellow Witch) make out with yours
Anonymous 01/15/25(Wed)19:58:30 No.103910959
1725719704795665
>>103910823
If you're using the "normal" scheduler, drop it, it's glitchy, go for "simple" instead
Anonymous 01/15/25(Wed)20:05:39 No.103911026
>>103910821
Based, I'm gonna make a green witch. LOOL we can be a coven LOOOL
Anonymous 01/15/25(Wed)20:06:08 No.103911032
>>103910821
this is a sorcerer, not a witch
Anonymous 01/15/25(Wed)20:09:18 No.103911066
00201-0
>>103910954
no objections here
Anonymous 01/15/25(Wed)20:16:32 No.103911150
>>103911066
>>103910954
You forgot to ask me. I object. Lesbianism is a sin.
Anonymous 01/15/25(Wed)20:17:36 No.103911160
>>103911150
Yellow Witch was born male
Anonymous 01/15/25(Wed)20:18:44 No.103911174
GW__00001_
Hello. I am Green Witch LOOL
Anonymous 01/15/25(Wed)20:23:09 No.103911224
>>103911174
Nice fries, some girl with curly blonde hair, a permanent angry expression, and a burger should show up to shit up the thread - I mean, share her burger with you.
Anonymous 01/15/25(Wed)20:24:48 No.103911240
1720689134657529
>>103910849
Thanks. Have some more 'eye' candy.
Anonymous 01/15/25(Wed)20:29:58 No.103911304
my fellow kinosovl poasters lend me your ear
Anonymous 01/15/25(Wed)20:31:15 No.103911317
Anonymous 01/15/25(Wed)20:34:18 No.103911356
ComfyUI_00022_
Anonymous 01/15/25(Wed)20:36:57 No.103911383
Anonymous 01/15/25(Wed)20:45:07 No.103911482
GW__00007_
>>103911356
LOOL. That's crazy anon.
Anonymous 01/15/25(Wed)20:48:33 No.103911519
snakeoil
This time it will be perfect
Anonymous 01/15/25(Wed)20:51:31 No.103911546
>>103911519
>Random crop.
Why not just smear shit on your dataset. It would be just as effective.
Anonymous 01/15/25(Wed)20:53:39 No.103911569
>>103911546
this, people who do crop deserve the rop
Anonymous 01/15/25(Wed)20:58:48 No.103911620
0
Anonymous 01/15/25(Wed)20:58:53 No.103911621
Anonymous 01/15/25(Wed)20:59:36 No.103911630
>>103911569
Even flip augmentation is arguably a terrible idea.
Anonymous 01/15/25(Wed)20:59:38 No.103911631
>>103911621
why the fuck would you crop a picture?
Anonymous 01/15/25(Wed)21:01:55 No.103911655
>>103911631
to reframe focus

plus to get rid of ai fuck ups
Anonymous 01/15/25(Wed)21:03:26 No.103911670
>>103911655
>Makes dataset
>Curates dataset
>Labels dataset
>Now let's have an algorithm randomly cut up segments of my images with no rhyme or reason.
Anonymous 01/15/25(Wed)21:04:59 No.103911695
>>103911655
wtf is wrong with you, why can't you train with real pictures, using AI shit to poison your model is always a shit idea
Anonymous 01/15/25(Wed)21:07:30 No.103911724
Anonymous 01/15/25(Wed)22:58:38 No.103912735
>>103903645
What is this?