/ldg/ - Local Diffusion General
Anonymous 01/14/25(Tue)13:50:28 | 311 comments | 83 images | 🔒 Locked
Discussion of Free and Open-Source Diffusion models.
What's in the Tank Edition
Last Bread: >>103886319
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/tdrussell/diffusion-pipe|https://github.com/kohya-ss/musubi-tuner
>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
What's in the Tank Edition
Last Bread: >>103886319
Newfrien Guide: https://rentry.org/ldg-lazy-getting
>UI
Metastable:https://metastable.studi
SwarmUI:https://github.com/mcmonkey
Forge:https://github.com/lllyasviel
reForge:https://github.com/Panchovi
ComfyUI:https://github.com/comfyano
InvokeAI:https://github.com/invoke-
>Models, LoRAs, & Upscalers
https://civitai.com
https://tensor.art/
https://openmodeldb.info
>Training
https://github.com/Nerogar/OneTrain
https://github.com/kohya-ss/sd-scri
https://github.com/derrian-distro/L
>HunyuanVideo
Comfy:https://github.com/kijai/Comf
Windows:https://rentry.org/crhcqq54
Training:https://github.com/tdrusse
>Flux
Forge:https://github.com/lllyasviel
ComfyUI:https://comfyanonymous.gith
DeDistilled:https://huggingface.co/
Pixelwave:https://civitai.com/model
>Misc
Share Metadata:https://catbox.moe|https:/
Img2Prompt:https://huggingface.co/s
Archive:https://rentry.org/sdg-link
Samplers:https://stable-diffusion-a
Digital Art Software:https://krita.org/en/
Txt2Img Plugin:https://kritaaidiffusion.com
Collagebaker:https://www.befunky.co
Video Collagebaker:https://kdenlive.org/e
>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 01/14/25(Tue)13:52:21 No.103893516
Blessed thread of frenship
Anonymous 01/14/25(Tue)13:53:29 No.103893535
>>103893516
nuh uh, ur gay and i'm gonna give you a swirlie, ya nerd
nuh uh, ur gay and i'm gonna give you a swirlie, ya nerd
Anonymous 01/14/25(Tue)13:53:52 No.103893538
watching
Anonymous 01/14/25(Tue)14:03:42 No.103893627
Calm thread of quietness
Anonymous 01/14/25(Tue)14:04:50 No.103893638
Getting there. Could really use some image to video.
Anonymous 01/14/25(Tue)14:11:39 No.103893703
Anonymous 01/14/25(Tue)14:12:37 No.103893711
>>103893703
Is the green guy inpainted at all? Really good details all around.
Is the green guy inpainted at all? Really good details all around.
Anonymous 01/14/25(Tue)14:14:59 No.103893731
>>103893711
gimp photobash, didn't inpaint anything
gimp photobash, didn't inpaint anything
Anonymous 01/14/25(Tue)14:22:13 No.103893803
Might be skill issue, but Flux can't do Rei Ayanami at all. I'm prompting things that would get me Rei in, say, Illustrious but Flux gives me someone that's very much not Rei.
Anonymous 01/14/25(Tue)14:25:44 No.103893849
>>103893803
why would you gen anime in flux when we have noob? can always use flux-o to img2img later on. just use a lora to hammer the concept into the model if you really want to use it.
why would you gen anime in flux when we have noob? can always use flux-o to img2img later on. just use a lora to hammer the concept into the model if you really want to use it.
Anonymous 01/14/25(Tue)14:26:39 No.103893862
Anonymous 01/14/25(Tue)14:28:20 No.103893884
>>103893849
Every time I try noob I get very weird lighting
Every time I try noob I get very weird lighting
Anonymous 01/14/25(Tue)14:28:38 No.103893887
is there any way to preview the whole video during the steps, rather than just the first frame
Anonymous 01/14/25(Tue)14:29:18 No.103893891
>>103893884
Post example?
Post example?
Anonymous 01/14/25(Tue)14:30:55 No.103893906
>>103893891
Blue board. Also I deleted the images in question.
So unironically 'trust me bro'. Again, might be skill issue on my part
Blue board. Also I deleted the images in question.
So unironically 'trust me bro'. Again, might be skill issue on my part
Anonymous 01/14/25(Tue)14:32:50 No.103893932
Anonymous 01/14/25(Tue)14:32:55 No.103893935
>>103893906 (Me)
Actually I do have an example, but still a blue board.
Actually I do have an example, but still a blue board.
Anonymous 01/14/25(Tue)14:33:35 No.103893944
>>103893935
post catbox so we can see your diaper porn
post catbox so we can see your diaper porn
Anonymous 01/14/25(Tue)14:34:09 No.103893949
Anonymous 01/14/25(Tue)14:38:28 No.103894002
>>103893731
nice
nice
Anonymous 01/14/25(Tue)14:38:59 No.103894017
>>103893949
>euler_ancestral
>cfg 8
Lower the CFG and use "euler_cfg_pp" sampler (not sure what it's called on AUTO). Noob tends to fry outputs like that if you're not careful. Some say rescaleCFG works but I haven't had much luck.
>euler_ancestral
>cfg 8
Lower the CFG and use "euler_cfg_pp" sampler (not sure what it's called on AUTO). Noob tends to fry outputs like that if you're not careful. Some say rescaleCFG works but I haven't had much luck.
Anonymous 01/14/25(Tue)14:39:56 No.103894028
>>103893906
noob isnt easy to prompt, it reacts to most tokens in a rather strong fashion. euler a cfg++ sampler, 1.5 cfg, sgm uniform scheduler, 'cfg rescale' (0.7 is fine) should get you started. really need to find the right way to prompt it. you can use one of those
https://civitai.com/models/971952/noobaiillustriousxl-stabilizer?modelVersionId=1265270
https://civitai.com/models/856285/pony-peoples-works?modelVersionId=1247042
at LOW strength to stabilize the model but it is more of a crutch.
noob isnt easy to prompt, it reacts to most tokens in a rather strong fashion. euler a cfg++ sampler, 1.5 cfg, sgm uniform scheduler, 'cfg rescale' (0.7 is fine) should get you started. really need to find the right way to prompt it. you can use one of those
https://civitai.com/models/971952/n
https://civitai.com/models/856285/p
at LOW strength to stabilize the model but it is more of a crutch.
Anonymous 01/14/25(Tue)14:40:18 No.103894033
https://files.catbox.moe/63a0ux.png
Anonymous 01/14/25(Tue)14:41:11 No.103894047
>>103894017
>>103894028
Go figure. I've been prompting on noob as if it was illustrious. Guess that was a mistake.
>>103894028
Go figure. I've been prompting on noob as if it was illustrious. Guess that was a mistake.
Anonymous 01/14/25(Tue)14:46:10 No.103894120
>>103894033
good bunni
good bunni
Anonymous 01/14/25(Tue)14:52:02 No.103894195
>>103893703
Cool.
Cool.
Anonymous 01/14/25(Tue)14:52:39 No.103894202
>>103894047
1girl, yoimiya \(genshin impact\) , genshin impact, millipen \(medium\), ;D, swimming, swimsuit, koi fish, evening, dim light, water surface, (dishwasher1910:0.4), masterpiece, best quality <lora:ponyv6_noobV1_2_adamW:0.4>
worst quality, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro
Steps: 28, Sampler: Euler Ancestral CFG++, Schedule type: SGM Uniform, CFG scale: 1.5
(and I think I had PAG active at 1.0 strength. adamW is 'people's works' lora)
1girl, yoimiya \(genshin impact\) , genshin impact, millipen \(medium\), ;D, swimming, swimsuit, koi fish, evening, dim light, water surface, (dishwasher1910:0.4), masterpiece, best quality <lora:ponyv6_noobV1_2_adamW:0.4>
worst quality, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro
Steps: 28, Sampler: Euler Ancestral CFG++, Schedule type: SGM Uniform, CFG scale: 1.5
(and I think I had PAG active at 1.0 strength. adamW is 'people's works' lora)
Anonymous 01/14/25(Tue)14:58:25 No.103894263
Anonymous 01/14/25(Tue)15:15:20 No.103894432
Anonymous 01/14/25(Tue)15:17:07 No.103894445
Now that the dust has settled which model is best
Anonymous 01/14/25(Tue)15:27:22 No.103894552
>>103894445
Hunyuan
Hunyuan
Anonymous 01/14/25(Tue)15:35:06 No.103894655
>>103894445
deep dream
deep dream
Anonymous 01/14/25(Tue)15:40:58 No.103894735
>>103894445
cardosAnimated_v10
cardosAnimated_v10
Anonymous 01/14/25(Tue)15:41:29 No.103894740
>>103894445
bstaber
bstaber
Anonymous 01/14/25(Tue)15:42:16 No.103894756
>>103894445
tunabiscuit
tunabiscuit
Anonymous 01/14/25(Tue)15:43:55 No.103894773
>>103894445
juggernautXL
juggernautXL
Anonymous 01/14/25(Tue)15:44:05 No.103894776
>>103893731
high effort
high effort
Anonymous 01/14/25(Tue)15:44:33 No.103894786
Anonymous 01/14/25(Tue)15:45:02 No.103894791
Anonymous 01/14/25(Tue)15:45:34 No.103894798
>>103894445
VGAN + CLIP
VGAN + CLIP
Anonymous 01/14/25(Tue)15:45:50 No.103894804
>>103894786
Cry
Cry
Anonymous 01/14/25(Tue)15:50:23 No.103894855
Which image generation models are best for training LoRAs on at the moment? Flux? SD3.5? Are there any advantages to using something older?
I want to train a LoRA to generate scenes from old 16-bit video games, with the proper top-down perspective and grid-aligned tiles and everything. I've been trying with Flux with some existing pixel art LoRAs, but most of the time I can't even get it to use the right perspective.
I want to train a LoRA to generate scenes from old 16-bit video games, with the proper top-down perspective and grid-aligned tiles and everything. I've been trying with Flux with some existing pixel art LoRAs, but most of the time I can't even get it to use the right perspective.
Anonymous 01/14/25(Tue)15:52:03 No.103894868
Anonymous 01/14/25(Tue)16:00:17 No.103894966
Anonymous 01/14/25(Tue)16:01:24 No.103894981
Anonymous 01/14/25(Tue)16:07:44 No.103895060
Anonymous 01/14/25(Tue)16:11:57 No.103895115
Been trying cosmos 7b in comfyui. I find it odd that generating images (single frame) works fine, and the full 1280x704x121 also works, but any frame length between those gives garbage results. Like, literally pure noise if you're much below 121 frames.
From the Cosmos paper:
>We employ a 3Dfactorized Rotary Position Embedding (RoPE) (Su et al., 2024) to allow the generation of arbitrary size, aspect ratio, and video length.
And they talk about how they also do joint image-video training. What's the deal then? Is this line from the paper just a lie? I would expect it to work like hunyuan where any frame length works.
From the Cosmos paper:
>We employ a 3Dfactorized Rotary Position Embedding (RoPE) (Su et al., 2024) to allow the generation of arbitrary size, aspect ratio, and video length.
And they talk about how they also do joint image-video training. What's the deal then? Is this line from the paper just a lie? I would expect it to work like hunyuan where any frame length works.
Anonymous 01/14/25(Tue)16:31:50 No.103895371
Anonymous 01/14/25(Tue)16:37:54 No.103895441
>>103893500
how did it go?
how did it go?
Anonymous 01/14/25(Tue)16:39:11 No.103895456
>>103895441
he dieded
he dieded
Anonymous 01/14/25(Tue)16:44:45 No.103895518
Anonymous 01/14/25(Tue)16:53:44 No.103895623
>adolf anime opening
Anonymous 01/14/25(Tue)16:55:20 No.103895645
what's the effect called where the focus shifts to the foreground and the background gets all blurry
Anonymous 01/14/25(Tue)16:55:46 No.103895650
>>103894445
SD3.5
SD3.5
Anonymous 01/14/25(Tue)17:03:29 No.103895738
>llmfags eating good
>imggenfags left out to dry
Why does this always happen?
>imggenfags left out to dry
Why does this always happen?
Anonymous 01/14/25(Tue)17:06:00 No.103895766
>>103895623
Cosmos?
Cosmos?
Anonymous 01/14/25(Tue)17:15:15 No.103895873
>>103895766
hunyuan, but it's Q6_K
hunyuan, but it's Q6_K
Anonymous 01/14/25(Tue)17:23:00 No.103895969
>>103895645
"blurred background" is a boorutag if youre using an anime model
"blurred background" is a boorutag if youre using an anime model
Anonymous 01/14/25(Tue)17:23:41 No.103895976
>>103895115
Ok, I've done some research and I think I can answer my own question. A dev on the github said cosmos only supports 121 frames. The paper talks about how they do progressive resizing of videos in stages during training, but also how they jointly do image-video training. So I think what happened, is that during training they trained on images, along with videos that progressively got higher res / longer. Ending with just a few aspect ratio buckets at certain resolutions. But images were trained on throughout. So the model can do images, and it can do 121 frame videos at one of the supported buckets, and that's it.
Kind of sucks, with hunyuan for experimenting you can do lower res and like 65 frames and it works fine, for much faster iteration. But you just can't with Cosmos.
Ok, I've done some research and I think I can answer my own question. A dev on the github said cosmos only supports 121 frames. The paper talks about how they do progressive resizing of videos in stages during training, but also how they jointly do image-video training. So I think what happened, is that during training they trained on images, along with videos that progressively got higher res / longer. Ending with just a few aspect ratio buckets at certain resolutions. But images were trained on throughout. So the model can do images, and it can do 121 frame videos at one of the supported buckets, and that's it.
Kind of sucks, with hunyuan for experimenting you can do lower res and like 65 frames and it works fine, for much faster iteration. But you just can't with Cosmos.
Anonymous 01/14/25(Tue)17:25:24 No.103896002
I stopped genning for a couple of months. I was using Pony back then. I came back to Illustrious being a thing and I'm trying to get the hang of it.
Is there a list of artist tags with example gens? I remember seeing on of these for Pony.
Is there a list of artist tags with example gens? I remember seeing on of these for Pony.
Anonymous 01/14/25(Tue)17:26:34 No.103896014
>>103896002
OP of /hdg/ should still have that fucked up spreadsheet of Illustrious artists
But this is a NoobAI general BTW
OP of /hdg/ should still have that fucked up spreadsheet of Illustrious artists
But this is a NoobAI general BTW
Anonymous 01/14/25(Tue)17:32:16 No.103896098
>>103896002
I'd actually recommend you to find illustrious/noob wildcards and try them first. The amount of styles is so monumental trying to pick one yourself from just a spreadsheet is a pain in the ass.
I'd actually recommend you to find illustrious/noob wildcards and try them first. The amount of styles is so monumental trying to pick one yourself from just a spreadsheet is a pain in the ass.
Anonymous 01/14/25(Tue)17:32:20 No.103896099
>>103894432
should she be bigger? cause she whalegirl
should she be bigger? cause she whalegirl
Anonymous 01/14/25(Tue)17:35:49 No.103896137
Anonymous 01/14/25(Tue)17:37:21 No.103896158
>>103896098
>find illustrious/noob wildcards
NTA where do you usually find wildcards like that? I don't think I see any on Civitai.
>find illustrious/noob wildcards
NTA where do you usually find wildcards like that? I don't think I see any on Civitai.
Anonymous 01/14/25(Tue)17:38:13 No.103896165
>>103896137
ugly bish
ugly bish
Anonymous 01/14/25(Tue)17:38:47 No.103896168
Anonymous 01/14/25(Tue)17:39:54 No.103896180
Hunyuan lora captioning is weird
A single word caption acting as an activator seems to be the only style that works well
A single word caption acting as an activator seems to be the only style that works well
Anonymous 01/14/25(Tue)17:41:53 No.103896204
>>103896014
My bad I came here from elsewhere and since I gen locally I thought it was the right place.
Thanks for the answer though!
>>103896098
I'll check it out!
>>103896137
This is great thanks!
My bad I came here from elsewhere and since I gen locally I thought it was the right place.
Thanks for the answer though!
>>103896098
I'll check it out!
>>103896137
This is great thanks!
Anonymous 01/14/25(Tue)17:42:06 No.103896207
>>103896180
nice i wonder how well the deepthroat et al loras work with a cartoon lora
nice i wonder how well the deepthroat et al loras work with a cartoon lora
Anonymous 01/14/25(Tue)17:44:21 No.103896237
>>103896158
I found it within two seconds of google search, anon.
https://huggingface.co/datasets/mimizukari/NoobAI-NAI-XL-Wildcards
I found it within two seconds of google search, anon.
https://huggingface.co/datasets/mim
Anonymous 01/14/25(Tue)17:49:02 No.103896296
Anonymous 01/14/25(Tue)17:50:01 No.103896308
Anonymous 01/14/25(Tue)17:53:38 No.103896351
>>103896296
this one is very good
this one is very good
Anonymous 01/14/25(Tue)17:54:32 No.103896364
>>103896308
>>103896351
Thanks. The background people, her upper body and runes (lightly). I forgot to img2img it at the end to clean up the inpainting artifacts so you can see the discontinuities a bit.
>>103896351
Thanks. The background people, her upper body and runes (lightly). I forgot to img2img it at the end to clean up the inpainting artifacts so you can see the discontinuities a bit.
GTX1060 6GB Resurrection 01/14/25(Tue)17:55:04 No.103896369
>>103894445
still AnimagineXL3.1
still AnimagineXL3.1
Anonymous 01/14/25(Tue)17:55:46 No.103896377
>>103894445
The answer is it depends on what you're trying to make.
The answer is it depends on what you're trying to make.
Anonymous 01/14/25(Tue)18:00:23 No.103896420
>>103896377
I don't think it does desu. My favorite model can do it all.
I don't think it does desu. My favorite model can do it all.
Anonymous 01/14/25(Tue)18:06:20 No.103896484
chinese new year
serious this time
serious this time
Anonymous 01/14/25(Tue)18:11:38 No.103896541
>>103896207
NSFW
https://files.catbox.moe/m9rkpl.mp4
Works okay.
Requires putting a lot of emphasis on being a cartoon in the prompt though, maybe my lora is undertrained
NSFW
https://files.catbox.moe/m9rkpl.mp4
Works okay.
Requires putting a lot of emphasis on being a cartoon in the prompt though, maybe my lora is undertrained
Anonymous 01/14/25(Tue)18:12:18 No.103896547
>reforge
what's wrong with forge?
what's wrong with forge?
Anonymous 01/14/25(Tue)18:14:09 No.103896571
>>103896547
I'm using Forge and there's nothing wrong with it
I'm using Forge and there's nothing wrong with it
Anonymous 01/14/25(Tue)18:18:30 No.103896618
>>103896571
then why does reforge exists you fucking retarded faggot retard, that's obviously what i'm asking
ffs
then why does reforge exists you fucking retarded faggot retard, that's obviously what i'm asking
ffs
Anonymous 01/14/25(Tue)18:24:10 No.103896672
Noob question.
Can I gen two exact same images at the same time except in the other one the character has eyes closed, and in the other one open? Otherwise they're the exact same.
Can I gen two exact same images at the same time except in the other one the character has eyes closed, and in the other one open? Otherwise they're the exact same.
Anonymous 01/14/25(Tue)18:25:09 No.103896685
>>103896672
lmao, these threads are amazing
lmao, these threads are amazing
Anonymous 01/14/25(Tue)18:26:25 No.103896702
>>103896685
Inpaiting in your friend.
Inpaiting in your friend.
Anonymous 01/14/25(Tue)18:26:52 No.103896707
>>103896672
Inpaint on top of your prompt, or use the same seed with a slightly modified prompt.
Inpaint on top of your prompt, or use the same seed with a slightly modified prompt.
Anonymous 01/14/25(Tue)18:28:15 No.103896727
Cool LORA.
https://civitai.com/models/1134895?modelVersionId=1276001
https://civitai.com/models/1134895?
Anonymous 01/14/25(Tue)18:29:21 No.103896743
Anonymous 01/14/25(Tue)18:31:16 No.103896766
Anonymous 01/14/25(Tue)18:31:56 No.103896778
>>103896541
NIce. I can finally coom to AI gens. I couldn't before because still images do nothing for me, but this is a game changer. Pray for me anon.
NIce. I can finally coom to AI gens. I couldn't before because still images do nothing for me, but this is a game changer. Pray for me anon.
Anonymous 01/14/25(Tue)18:33:40 No.103896797
>>103896685
should we lambast newfrens or help them out? i always go back and forth
should we lambast newfrens or help them out? i always go back and forth
Anonymous 01/14/25(Tue)18:34:43 No.103896803
>>103896797
it doesn't matter, these threads basically destroyed my faith in humanity
it doesn't matter, these threads basically destroyed my faith in humanity
Anonymous 01/14/25(Tue)18:38:16 No.103896846
How big is your slop folder?
Anonymous 01/14/25(Tue)18:38:18 No.103896848
This is what happens when no new models release
Fucking gay
Fucking gay
Anonymous 01/14/25(Tue)18:38:51 No.103896855
>>103896803
You are very smart, probably there isn't a single subject in the entire you are a noob about. The humanity is net positive thanks to people like you.
You are very smart, probably there isn't a single subject in the entire you are a noob about. The humanity is net positive thanks to people like you.
Anonymous 01/14/25(Tue)18:38:54 No.103896857
>>103896848
We just got cosmos
We just got cosmos
Anonymous 01/14/25(Tue)18:39:29 No.103896864
>>103896204
>since I gen locally I thought it was the right place.
localchads together strong
>>103896857
IMAGE MODELS !!
>since I gen locally I thought it was the right place.
localchads together strong
>>103896857
IMAGE MODELS !!
Anonymous 01/14/25(Tue)18:41:00 No.103896886
Anonymous 01/14/25(Tue)18:43:02 No.103896906
>>103896855
i'm better at this than anyone here, but that's not the issue
i've been here since dec 2023, and it's been an awful experience every single time
it's truly amazing how every single question is either ignored or wrongly answered on purpose
i would literally kill every single person itt if i could
i'm better at this than anyone here, but that's not the issue
i've been here since dec 2023, and it's been an awful experience every single time
it's truly amazing how every single question is either ignored or wrongly answered on purpose
i would literally kill every single person itt if i could
Anonymous 01/14/25(Tue)18:44:56 No.103896935
Whats the difference between this and stable diffusion? I've been living under a rock.
Anonymous 01/14/25(Tue)18:45:30 No.103896939
>>103896935
what is 'this'
what is 'this'
Anonymous 01/14/25(Tue)18:46:08 No.103896946
>>103896935
"local models" is a broader term that also envelops "stable diffusion".
"local models" is a broader term that also envelops "stable diffusion".
Anonymous 01/14/25(Tue)18:46:50 No.103896956
Anonymous 01/14/25(Tue)18:47:05 No.103896958
>>103896906
See, as said, very very very smart. Thanks for browsing 4chan.
See, as said, very very very smart. Thanks for browsing 4chan.
Anonymous 01/14/25(Tue)18:47:36 No.103896966
We're half way through January and I don't see img2vid. Worried yet, "director"?
Anonymous 01/14/25(Tue)18:47:40 No.103896968
>>103896906
>i would literally kill every single person itt if i could
I still love you anon *smooch*
>>103896956
you should post this again because i missed it the other times you posted it
>i would literally kill every single person itt if i could
I still love you anon *smooch*
>>103896956
you should post this again because i missed it the other times you posted it
Anonymous 01/14/25(Tue)18:48:13 No.103896976
the director isn't real he can't hurt you
Anonymous 01/14/25(Tue)18:48:33 No.103896980
>>103896956
don't care if you believe it or not
>>103896958
>>103896968
see, two subhuman imbeciles with nothing to say
don't care if you believe it or not
>>103896958
>>103896968
see, two subhuman imbeciles with nothing to say
Anonymous 01/14/25(Tue)18:49:00 No.103896984
This is a nice thread
Anonymous 01/14/25(Tue)18:51:50 No.103897025
>>103896984
you too
you too
Anonymous 01/14/25(Tue)18:52:04 No.103897027
>i've been here since dec 2023
newfag
newfag
Anonymous 01/14/25(Tue)18:53:17 No.103897037
>i'm funny
Anonymous 01/14/25(Tue)18:55:14 No.103897073
>>103896886
Badass gen
Badass gen
Anonymous 01/14/25(Tue)18:56:01 No.103897080
>>103896939
Local diffusion. Are the gens really separated by who runs stuff locally, vs not?
Local diffusion. Are the gens really separated by who runs stuff locally, vs not?
Anonymous 01/14/25(Tue)18:56:12 No.103897082
>>103897073
Thanks!
Thanks!
Anonymous 01/14/25(Tue)18:58:19 No.103897105
>>103897080
yeah there're a bunch of dalle (paid service) specific threads on this board and others
yeah there're a bunch of dalle (paid service) specific threads on this board and others
Anonymous 01/14/25(Tue)19:00:01 No.103897127
>>103897080
No. /sdg/ is a place where people play with their images, roleplay as the people in their images and just post their generations with very little consideration to anything behind it.
/ldg/ is more about the discussion of the technology, trying and discussing new tools for diffusion models and posting the results of their experiments.
Unfortunately the distinct personalities of /sdg/ and /ldg/ do not get along well in the same thread so the split was necessitated.
No. /sdg/ is a place where people play with their images, roleplay as the people in their images and just post their generations with very little consideration to anything behind it.
/ldg/ is more about the discussion of the technology, trying and discussing new tools for diffusion models and posting the results of their experiments.
Unfortunately the distinct personalities of /sdg/ and /ldg/ do not get along well in the same thread so the split was necessitated.
Anonymous 01/14/25(Tue)19:02:42 No.103897162
>>103897127
>Unfortunately the distinct personalities of /sdg/ and /ldg/ do not get along well in the same thread so the split was necessitated.
he means the pathetic talent-less troons that reside 24/7 in both of these worthless shithole threads btw
>Unfortunately the distinct personalities of /sdg/ and /ldg/ do not get along well in the same thread so the split was necessitated.
he means the pathetic talent-less troons that reside 24/7 in both of these worthless shithole threads btw
Anonymous 01/14/25(Tue)19:03:47 No.103897173
>>103897162
Something on your mind, anon? Let it all out. Big tears.
Something on your mind, anon? Let it all out. Big tears.
Anonymous 01/14/25(Tue)19:04:12 No.103897181
>>103897127
If you're not fucking with me...christ.
>>103897162
Gathered as much.
Well, thanks for the info and OP resources.
If you're not fucking with me...christ.
>>103897162
Gathered as much.
Well, thanks for the info and OP resources.
Anonymous 01/14/25(Tue)19:05:51 No.103897202
Anonymous 01/14/25(Tue)19:07:20 No.103897221
Anonymous 01/14/25(Tue)19:07:48 No.103897225
I've worked with generative synthesis pre Google's Deepdream, ask me anything
Anonymous 01/14/25(Tue)19:07:55 No.103897227
>>103897221
case and point
case and point
Anonymous 01/14/25(Tue)19:08:51 No.103897241
Anonymous 01/14/25(Tue)19:09:55 No.103897257
why do anons come in here just to cry? i dont understand
Anonymous 01/14/25(Tue)19:12:19 No.103897285
>>103897257
There are only two reasons to be here.
1. Something really exciting happened and we want to talk about it.
2. Nothing is happening, we are bored and want to express that.
There are only two reasons to be here.
1. Something really exciting happened and we want to talk about it.
2. Nothing is happening, we are bored and want to express that.
Anonymous 01/14/25(Tue)19:12:46 No.103897295
>>103897257
no one is crying you subhuman troon
i come here every once in a while to ask something, just to see what you subhuman worthless losers have to say about it, and insult you for fun for a few posts because you deserve it
the only one crying here is you because my posts offends you, because you know everything i say is true
no one is crying you subhuman troon
i come here every once in a while to ask something, just to see what you subhuman worthless losers have to say about it, and insult you for fun for a few posts because you deserve it
the only one crying here is you because my posts offends you, because you know everything i say is true
Anonymous 01/14/25(Tue)19:14:48 No.103897311
>>103897285
>Nothing is happening, we are bored and want to express that.
good point. image generation anons are starving right now
>Nothing is happening, we are bored and want to express that.
good point. image generation anons are starving right now
Anonymous 01/14/25(Tue)19:15:49 No.103897326
Anonymous 01/14/25(Tue)19:15:57 No.103897328
>>103897241
Sorry I'm not Chinese
Sorry I'm not Chinese
Anonymous 01/14/25(Tue)19:17:35 No.103897345
>>103897225
how did google got so good on video gen? their veo2 model is fucking amazing
https://www.youtube.com/watch?v=w_1abtGULtg
how did google got so good on video gen? their veo2 model is fucking amazing
https://www.youtube.com/watch?v=w_1
Anonymous 01/14/25(Tue)19:19:52 No.103897371
>>103897345
Good data. But you don't need to be a researcher to guess that.
Good data. But you don't need to be a researcher to guess that.
Anonymous 01/14/25(Tue)19:22:35 No.103897396
>>103897371
so basically everyone can be an AI researcher because the only thing that matter is good data? that's dissapointing
so basically everyone can be an AI researcher because the only thing that matter is good data? that's dissapointing
Anonymous 01/14/25(Tue)19:26:56 No.103897431
Debo got triggered because anon asked him if he had a job followed by asking him if he was on disability after dodging the question.
Expect him to lash out
Expect him to lash out
Anonymous 01/14/25(Tue)19:26:56 No.103897432
>>103897396
I'm beginning to doubt his qualifications.
I'm beginning to doubt his qualifications.
Anonymous 01/14/25(Tue)19:28:03 No.103897444
>>103896966
I worry for my vramchad brothers. We are all localfags... we live and die together.
I worry for my vramchad brothers. We are all localfags... we live and die together.
Anonymous 01/14/25(Tue)19:29:50 No.103897464
Anonymous 01/14/25(Tue)19:30:03 No.103897466
https://www.reddit.com/r/StableDiffusion/comments/1i19op9/comment/m74m55c/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button
>We haven't launched Juggernaut Flux yet it is still in the testing phase.
Maybe a good finetune on Flux is on the road
>We haven't launched Juggernaut Flux yet it is still in the testing phase.
Maybe a good finetune on Flux is on the road
Anonymous 01/14/25(Tue)19:31:33 No.103897479
>>103897466
i hope it looks very similar style-wise to the XL version because we need more gens like that
i hope it looks very similar style-wise to the XL version because we need more gens like that
Anonymous 01/14/25(Tue)19:31:48 No.103897482
>>103897466
Juggernaut XL is one of the sloppiest model out there
Juggernaut XL is one of the sloppiest model out there
Anonymous 01/14/25(Tue)19:32:13 No.103897486
>>103896618
reforge has a wider selection of samplers and schedulers and does not shit the bed when you try this thing called 'inpainting'. forge can run flux.
>>103896886
very nice man, may I suggest a sliver of contrast?
reforge has a wider selection of samplers and schedulers and does not shit the bed when you try this thing called 'inpainting'. forge can run flux.
>>103896886
very nice man, may I suggest a sliver of contrast?
Anonymous 01/14/25(Tue)19:33:20 No.103897500
>>103897432
You're saying I'm wrong?
You're saying I'm wrong?
Anonymous 01/14/25(Tue)19:34:21 No.103897512
Anonymous 01/14/25(Tue)19:35:11 No.103897520
>>103897512
I'd be okay with the government restricting access to AI for people like this.
I'd be okay with the government restricting access to AI for people like this.
Anonymous 01/14/25(Tue)19:35:55 No.103897525
>>103897520
kek
kek
Anonymous 01/14/25(Tue)19:38:24 No.103897544
>>103897512
I am now wowed by what I am seeing here.
I am now wowed by what I am seeing here.
Anonymous 01/14/25(Tue)19:39:25 No.103897553
Anonymous 01/14/25(Tue)19:40:21 No.103897560
>>103897466
>it is still in the testing phase.
so theres still time for them to restart from scratch and not make it look like that
>it is still in the testing phase.
so theres still time for them to restart from scratch and not make it look like that
Anonymous 01/14/25(Tue)19:40:34 No.103897563
>>103897486
well, forge's never shit the bed when inpainting, desu
that being said, i don't really use forge because the image quality is lower than A1111, no idea why
well, forge's never shit the bed when inpainting, desu
that being said, i don't really use forge because the image quality is lower than A1111, no idea why
Anonymous 01/14/25(Tue)19:41:45 No.103897577
>>103897512
I don't get it, those picture could've been made by base Flux easily, what are they trying to improve?
I don't get it, those picture could've been made by base Flux easily, what are they trying to improve?
Anonymous 01/14/25(Tue)19:43:42 No.103897601
>>103897577
To give the devil it's due, I assume they're working with base Flux and not dedistilled. Impossible to get a strong tune on that thing.
To give the devil it's due, I assume they're working with base Flux and not dedistilled. Impossible to get a strong tune on that thing.
Anonymous 01/14/25(Tue)19:44:47 No.103897615
>>103897601
>I assume they're working with base Flux and not dedistilled.
if it's true then they're more retarded than expected, dedistilled was created so that we can finetune that mf
>I assume they're working with base Flux and not dedistilled.
if it's true then they're more retarded than expected, dedistilled was created so that we can finetune that mf
Anonymous 01/14/25(Tue)19:44:47 No.103897616
>>103897577
If it's Flux + tits - buttchin that'll be good enough I suppose
If it's Flux + tits - buttchin that'll be good enough I suppose
Anonymous 01/14/25(Tue)19:45:47 No.103897625
>>103897616
juggernaut is good at nfsw?
juggernaut is good at nfsw?
Anonymous 01/14/25(Tue)19:45:59 No.103897628
Anonymous 01/14/25(Tue)19:46:15 No.103897630
>>103897616
I can clearly see a butt on that asian woman's chin.
I can clearly see a butt on that asian woman's chin.
Anonymous 01/14/25(Tue)19:49:34 No.103897654
>>103897563
that really should not happen tho.
>>103897577
the asian girl, meh. here early flux shit for reference
that really should not happen tho.
>>103897577
the asian girl, meh. here early flux shit for reference
Anonymous 01/14/25(Tue)19:54:27 No.103897698
Anonymous 01/14/25(Tue)20:00:11 No.103897753
>>103897563
>that being said, i don't really use forge
Yeah, gradio based UIs should only be a tempo-
>because the image quality is lower than A1111
Jesus
How do Gradiots still exist in 2025?
>that being said, i don't really use forge
Yeah, gradio based UIs should only be a tempo-
>because the image quality is lower than A1111
Jesus
How do Gradiots still exist in 2025?
Anonymous 01/14/25(Tue)20:02:34 No.103897783
>>103897698
hm. base gens or upscaling? there are a lot of commandline args that might.. do something.
if I were you I would not rest until thats resolved or just.. stop caring because a1111 is just not the thing anymore.
I kinda stayed on comfy after flux and just recently pulled both forge and reforge again. think I'm gonna keep reforge. hasta la vista, forgeman.
hm. base gens or upscaling? there are a lot of commandline args that might.. do something.
if I were you I would not rest until thats resolved or just.. stop caring because a1111 is just not the thing anymore.
I kinda stayed on comfy after flux and just recently pulled both forge and reforge again. think I'm gonna keep reforge. hasta la vista, forgeman.
Anonymous 01/14/25(Tue)20:02:41 No.103897784
>>103897753
you wouldn't understand comfytroon
keep connecting 34 little nodes to do basic PS shit, lmao
you wouldn't understand comfytroon
keep connecting 34 little nodes to do basic PS shit, lmao
Anonymous 01/14/25(Tue)20:05:26 No.103897812
>>103897783
>base gens or upscaling?
base, only noticeable for (hyper)realism, tho, anime trash looks the same
>if I were you
i don't really give a shit, all these UIs are trash, so I just use whatever gets the job done
>base gens or upscaling?
base, only noticeable for (hyper)realism, tho, anime trash looks the same
>if I were you
i don't really give a shit, all these UIs are trash, so I just use whatever gets the job done
Anonymous 01/14/25(Tue)20:11:39 No.103897870
Will this matter at all for us?
Anonymous 01/14/25(Tue)20:12:41 No.103897879
>>103897870
it's just some video game scam techniques like they did with DLSS
it's just some video game scam techniques like they did with DLSS
Anonymous 01/14/25(Tue)20:13:53 No.103897893
>>103897870
Not for AI stuff
Not for AI stuff
Anonymous 01/14/25(Tue)20:14:38 No.103897900
>>103897812
I see, anime trash, ouch lol. but seriously, your choice of upscaler, scheduler, the amount of noise you inject into the resampling process etc matter a lot more - and yeah they are all trash in their own way.
>>103897893
nice stuff wizardman
I see, anime trash, ouch lol. but seriously, your choice of upscaler, scheduler, the amount of noise you inject into the resampling process etc matter a lot more - and yeah they are all trash in their own way.
>>103897893
nice stuff wizardman
Anonymous 01/14/25(Tue)20:15:09 No.103897909
Do you guys use teacache for HY?
Anonymous 01/14/25(Tue)20:16:55 No.103897928
>>103897900
I know, I matched everything, and I mean everything, without any upscaling, I mean the base image
the quality is not the same
I know, I matched everything, and I mean everything, without any upscaling, I mean the base image
the quality is not the same
Anonymous 01/14/25(Tue)20:17:04 No.103897932
>>103897909
yeah I do, I'm using that one at 0.10, works really well
https://github.com/welltop-cn/ComfyUI-TeaCache
now I can do a 512x512x129f in 7 min, which is quite insane, we're talking about a 13b model here, but with the SageAttention + TeaCache + the fact it's a distilled model (it only works at cfg = 1 -> 2x faster), running hunyuan seems bearable now
yeah I do, I'm using that one at 0.10, works really well
https://github.com/welltop-cn/Comfy
now I can do a 512x512x129f in 7 min, which is quite insane, we're talking about a 13b model here, but with the SageAttention + TeaCache + the fact it's a distilled model (it only works at cfg = 1 -> 2x faster), running hunyuan seems bearable now
Anonymous 01/14/25(Tue)20:18:12 No.103897942
>>103897909
Yeah. 0.07 for me
Yeah. 0.07 for me
Anonymous 01/14/25(Tue)20:26:20 No.103898012
Uhhh sirs it's recommended to use 0.15 for HyVid
Anonymous 01/14/25(Tue)20:28:05 No.103898023
>>103898012
the quality isn't great at 0.15, try it you'll see saar
the quality isn't great at 0.15, try it you'll see saar
Anonymous 01/14/25(Tue)20:28:17 No.103898025
>>103898012
And it looks like shit
And it looks like shit
Anonymous 01/14/25(Tue)20:30:49 No.103898056
>>103898023
>>103898025
Do you really think so? I tried testing other values and thought 0.15 looked best.
>>103898025
Do you really think so? I tried testing other values and thought 0.15 looked best.
Anonymous 01/14/25(Tue)20:31:29 No.103898063
>>103897577
this is like saying that comfyui is bad for generating images because the author is bad at genning
this is like saying that comfyui is bad for generating images because the author is bad at genning
Anonymous 01/14/25(Tue)20:32:50 No.103898078
>>103898056
>0.15 looked best
how can it look the best, the more you improve that number, the worse the quality gets, it'll always look the best at 0 when there's no compression
>0.15 looked best
how can it look the best, the more you improve that number, the worse the quality gets, it'll always look the best at 0 when there's no compression
Anonymous 01/14/25(Tue)20:32:57 No.103898080
Anonymous 01/14/25(Tue)20:33:54 No.103898085
>>103898063
you would assume that the guy training a model would at least shows what's more appealing about his finetune compared to the base model no?
you would assume that the guy training a model would at least shows what's more appealing about his finetune compared to the base model no?
Anonymous 01/14/25(Tue)20:34:11 No.103898088
>>103898056
https://files.catbox.moe/85qez5.mp4
Off - 0.07
0.15 - 0.30
Not the best example here but the quality loss becomes much more evident in videos with lots of movement
https://files.catbox.moe/85qez5.mp4
Off - 0.07
0.15 - 0.30
Not the best example here but the quality loss becomes much more evident in videos with lots of movement
Anonymous 01/14/25(Tue)20:34:37 No.103898093
Anonymous 01/14/25(Tue)20:35:43 No.103898105
Anonymous 01/14/25(Tue)20:37:59 No.103898131
>>103898088
yeah, when the character starts to tremble for no reason you know you've put a too high value
yeah, when the character starts to tremble for no reason you know you've put a too high value
Anonymous 01/14/25(Tue)20:39:43 No.103898147
>>103898088
hmmm I may have to redo my series on old ladies eating hotdogs now with lower compression
hmmm I may have to redo my series on old ladies eating hotdogs now with lower compression
Anonymous 01/14/25(Tue)20:42:44 No.103898179
>>103898085
your assumption is unfortunately incorrect
your assumption is unfortunately incorrect
Anonymous 01/14/25(Tue)20:44:39 No.103898201
>>103898179
you're right, I'm expecting them to not act like retards when showcasing the capabilities of a model, I guess I overestimate their intelligence quite a bit here, that won't happen again
you're right, I'm expecting them to not act like retards when showcasing the capabilities of a model, I guess I overestimate their intelligence quite a bit here, that won't happen again
Anonymous 01/14/25(Tue)20:46:02 No.103898219
Anonymous 01/14/25(Tue)20:46:17 No.103898222
>>103898085
to me its like asking comfy to be good at genning, the guy obviously is pretty busy with this repo all day and doesn't have time (like other anons do) to generate images all day, juggernaut's author probably spends most of his time finetuning shit, they both need a guy or several to generate images for them
to me its like asking comfy to be good at genning, the guy obviously is pretty busy with this repo all day and doesn't have time (like other anons do) to generate images all day, juggernaut's author probably spends most of his time finetuning shit, they both need a guy or several to generate images for them
Anonymous 01/14/25(Tue)20:46:26 No.103898225
When you realize a good number of crypto bros jumped ship and morphed into AI bros you can begin to account for some of the retardation you see in the space.
Anonymous 01/14/25(Tue)20:47:34 No.103898237
>>103898219
pixelwave flux
pixelwave flux
Anonymous 01/14/25(Tue)20:47:35 No.103898238
>>103898222
but juggernaut looks like shit tho
>>103898225
you can always tell whos in it for the love of the game and whos just trying to stoke their own ego
but juggernaut looks like shit tho
>>103898225
you can always tell whos in it for the love of the game and whos just trying to stoke their own ego
Anonymous 01/14/25(Tue)20:50:35 No.103898267
Juggernaut (along with Dreamshaper) are to blame for the overly shiny plastic synthetic slop look that's so prevalent in models. They're the last people I would trust with a Flux finetune
Anonymous 01/14/25(Tue)20:50:40 No.103898271
>>103898238
>whos in it for the love of the game and whos just trying to stoke their own ego
Which one is juggernut. And which one am I?
>whos in it for the love of the game and whos just trying to stoke their own ego
Which one is juggernut. And which one am I?
Anonymous 01/14/25(Tue)20:51:50 No.103898289
>>103898237
Thganks.
Thganks.
Anonymous 01/14/25(Tue)20:53:53 No.103898307
>>103898238
when zavy or the crystal clear guys make something flux, I'll be all ears. we'll see.. let them upgrade their ish.
when zavy or the crystal clear guys make something flux, I'll be all ears. we'll see.. let them upgrade their ish.
Anonymous 01/14/25(Tue)20:54:10 No.103898309
Anyone tried this? https://github.com/kohya-ss/musubi-tuner
Anonymous 01/14/25(Tue)20:55:21 No.103898314
Anonymous 01/14/25(Tue)20:58:55 No.103898346
>>103898309
No but it's Kohya so I've heard it does and it's Kohya, aka the one man in Japan who knows anything about ML.
No but it's Kohya so I've heard it does and it's Kohya, aka the one man in Japan who knows anything about ML.
Anonymous 01/14/25(Tue)20:59:49 No.103898356
>>103898225
>a good number of crypto bros jumped ship and morphed into AI bros
Detrimental. No wonder they think we're all scammers.
>a good number of crypto bros jumped ship and morphed into AI bros
Detrimental. No wonder they think we're all scammers.
Anonymous 01/14/25(Tue)21:01:19 No.103898376
>>103898309
Yeah made a hunyuan lora with 12gb vram
Yeah made a hunyuan lora with 12gb vram
Anonymous 01/14/25(Tue)21:01:38 No.103898379
>>103898346
I'm curious if there are any noticible quality differences between LoRAs trained with the kohya script and the diffusion-pipe script.
I'm curious if there are any noticible quality differences between LoRAs trained with the kohya script and the diffusion-pipe script.
Anonymous 01/14/25(Tue)21:02:15 No.103898386
>>103898309
Yeah, I've trained a few loras with it, you'll know what to expect if you're familiar with sd-scripts. Diffusion-pipe was just a nightmare trying to get running so I'm glad Kohya came along and provided an alternative
Yeah, I've trained a few loras with it, you'll know what to expect if you're familiar with sd-scripts. Diffusion-pipe was just a nightmare trying to get running so I'm glad Kohya came along and provided an alternative
Anonymous 01/14/25(Tue)21:05:41 No.103898431
>>103898376
What resolution did you train at?
What resolution did you train at?
Anonymous 01/14/25(Tue)21:05:45 No.103898434
>>103898379
>quality differences
I haven't tried both, only hit the pipe but I'm gonna go out on a limb and say any gains you'd see in one over the other will be miniscule compared to the impact the dataset itself has.
Both otherwise have the same feature set and are built off the same stuff with more parameter differences.
>quality differences
I haven't tried both, only hit the pipe but I'm gonna go out on a limb and say any gains you'd see in one over the other will be miniscule compared to the impact the dataset itself has.
Both otherwise have the same feature set and are built off the same stuff with more parameter differences.
Anonymous 01/14/25(Tue)21:06:45 No.103898452
>>103898431
64x64
64x64
Anonymous 01/14/25(Tue)21:09:37 No.103898482
does v-pred work with normal ctrolnet?
Anonymous 01/14/25(Tue)21:12:25 No.103898513
Anonymous 01/14/25(Tue)21:13:05 No.103898519
>Flux released five months ago
>only one finetune
>only one finetune
Anonymous 01/14/25(Tue)21:14:18 No.103898529
Anonymous 01/14/25(Tue)21:14:42 No.103898533
>>103898519
there will be more when the 5090 will be released, 32gb of vram is enough for a real finetune of Flux
there will be more when the 5090 will be released, 32gb of vram is enough for a real finetune of Flux
Anonymous 01/14/25(Tue)21:23:15 No.103898618
Anonymous 01/14/25(Tue)21:30:31 No.103898668
Anonymous 01/14/25(Tue)21:34:10 No.103898700
Anonymous 01/14/25(Tue)21:37:54 No.103898734
this is what flux thinks "unreal tournament" is
Anonymous 01/14/25(Tue)21:43:02 No.103898778
>>103898700
Airhead phenotype
Airhead phenotype
Anonymous 01/14/25(Tue)21:51:48 No.103898843
Anonymous 01/14/25(Tue)21:56:24 No.103898877
Anonymous 01/14/25(Tue)22:08:23 No.103898983
Anonymous 01/14/25(Tue)22:12:06 No.103899018
>>103898983
not bad but both images have small and easy to fix mistakes. last one missing a tooth, one before got some stocking weirdness. it's all about the illusion and errors interfere.
not bad but both images have small and easy to fix mistakes. last one missing a tooth, one before got some stocking weirdness. it's all about the illusion and errors interfere.
Anonymous 01/14/25(Tue)22:22:33 No.103899113
Anonymous 01/14/25(Tue)22:28:56 No.103899175
Anonymous 01/14/25(Tue)22:29:33 No.103899180
>>103898778
The best kind.
The best kind.
Anonymous 01/14/25(Tue)22:30:41 No.103899191
>>103899018
I find it funny when people demand perfection from ai when its imitating humans which aren't perfect either. I have seen images from real people and thought: "damn if that hand looked like that in one of gens I would definitely be conscious about it and try to inpaint it or something"
I find it funny when people demand perfection from ai when its imitating humans which aren't perfect either. I have seen images from real people and thought: "damn if that hand looked like that in one of gens I would definitely be conscious about it and try to inpaint it or something"
Anonymous 01/14/25(Tue)22:33:13 No.103899213
>>103899191
that was a critique on you, not AI
that was a critique on you, not AI
Anonymous 01/14/25(Tue)22:38:01 No.103899252
Anonymous 01/14/25(Tue)22:40:09 No.103899267
>>103899191
I know what you mean, I looked at lots of images and focused on the hands specifically a while ago and had the exact same thought. ugly and miscolored hands are quite common actually. still,
I know what you mean, I looked at lots of images and focused on the hands specifically a while ago and had the exact same thought. ugly and miscolored hands are quite common actually. still,
Anonymous 01/14/25(Tue)22:40:24 No.103899273
https://www.youtube.com/watch?v=biQew3V1Dv4
this is why we definitely need the hunyuan i2v model, i2v is the sovl of videogen
this is why we definitely need the hunyuan i2v model, i2v is the sovl of videogen
Anonymous 01/14/25(Tue)22:40:38 No.103899277
>>103899191
>demand perfection
its not about perfection, the other anon is right. you are producing lazy slop.
>demand perfection
its not about perfection, the other anon is right. you are producing lazy slop.
Anonymous 01/14/25(Tue)22:41:38 No.103899283
>>103897466
Finally, a fine-tune that has the potential to be big enough to bury the idea that FLUX can be modified in a significant way.
Finally, a fine-tune that has the potential to be big enough to bury the idea that FLUX can be modified in a significant way.
Anonymous 01/14/25(Tue)22:42:10 No.103899289
Anonymous 01/14/25(Tue)22:42:46 No.103899293
I am a top 1% poster on the stable diffusion subreddit and I certify this thread SLOP
Anonymous 01/14/25(Tue)22:44:10 No.103899300
>>103899293
they have castes on reddit? jeez, and I thought we were bad
they have castes on reddit? jeez, and I thought we were bad
Anonymous 01/14/25(Tue)22:45:52 No.103899314
>>103899300
>they have castes on reddit?
they always had, when someone says too much controvertial stuff according to libtards on the reddit like "there's 2 genders", everytime that guy makes a comment it'll be hidden and you'll have to click on it to see what he has to say, we can say whatever we want on 4chan but at least every voice can be heard without any biases
>they have castes on reddit?
they always had, when someone says too much controvertial stuff according to libtards on the reddit like "there's 2 genders", everytime that guy makes a comment it'll be hidden and you'll have to click on it to see what he has to say, we can say whatever we want on 4chan but at least every voice can be heard without any biases
Anonymous 01/14/25(Tue)22:48:46 No.103899331
>>103899277
right how exactly? by some anon subjective opinion? if a person has misaligned teeth I don't care. If you think my slop is lazy thats on you, downvote me all you want
right how exactly? by some anon subjective opinion? if a person has misaligned teeth I don't care. If you think my slop is lazy thats on you, downvote me all you want
Anonymous 01/14/25(Tue)22:50:07 No.103899338
>>103899331
stop defending your lazy slop, instead try to get better.
stop defending your lazy slop, instead try to get better.
Anonymous 01/14/25(Tue)22:50:50 No.103899341
Anonymous 01/14/25(Tue)22:51:19 No.103899344
kek, another win for the AI, if it makes funny situations like that I'm all for it
Anonymous 01/14/25(Tue)22:51:20 No.103899345
>>103899331
>by some anon subjective opinion?
I really hope you're not about to...
>if a person has misaligned teeth I don't care.
oh... goddamit man.
>by some anon subjective opinion?
I really hope you're not about to...
>if a person has misaligned teeth I don't care.
oh... goddamit man.
Anonymous 01/14/25(Tue)22:51:31 No.103899346
>>103899331
The critique of your images was unnecessary even if true, but your inability to handle that critique makes you the loser in this argument.
The critique of your images was unnecessary even if true, but your inability to handle that critique makes you the loser in this argument.
Anonymous 01/14/25(Tue)22:52:36 No.103899355
>>103899344
why would i want to read that block of text from other social media about a story from yesterdays news
why would i want to read that block of text from other social media about a story from yesterdays news
Anonymous 01/14/25(Tue)22:53:35 No.103899364
>>103899355
I'll give you the tl:dr, someone used AI to pass himself as Brad Pitt and managed to scam fucking 800k dollars to some rich roastie kek
I'll give you the tl:dr, someone used AI to pass himself as Brad Pitt and managed to scam fucking 800k dollars to some rich roastie kek
Anonymous 01/14/25(Tue)22:54:02 No.103899367
>>103899355
tldr Angelina Jolie froze Brad Bitt's assets needed to treat kidney cancer and the helpful woman transferred him the 830k euros.
tldr Angelina Jolie froze Brad Bitt's assets needed to treat kidney cancer and the helpful woman transferred him the 830k euros.
Anonymous 01/14/25(Tue)22:54:20 No.103899372
Anonymous 01/14/25(Tue)22:54:50 No.103899377
>>103899372
why?
why?
Anonymous 01/14/25(Tue)22:55:05 No.103899378
>>103899344
>all the money she had left after her divorce from her millionaire husband
my heart bleeds
>all the money she had left after her divorce from her millionaire husband
my heart bleeds
Anonymous 01/14/25(Tue)22:55:21 No.103899384
Anonymous 01/14/25(Tue)22:56:06 No.103899391
>>103899378
so it wasn't really her money, she stole from her husband and "Brad Pitt" took it back from her, beautiful poetry if you ask me
so it wasn't really her money, she stole from her husband and "Brad Pitt" took it back from her, beautiful poetry if you ask me
Anonymous 01/14/25(Tue)22:57:39 No.103899407
>>103899391
now this story will make sure she never needs another penny, she'll be fine
now this story will make sure she never needs another penny, she'll be fine
Anonymous 01/14/25(Tue)22:57:45 No.103899410
Anonymous 01/14/25(Tue)22:59:45 No.103899434
>>103899344
maybe that's an unpopular opinion, but if you're that retarded to fall to such an obvious scam, you don't deserve to be protected by the police, I call that natural selection at this point
maybe that's an unpopular opinion, but if you're that retarded to fall to such an obvious scam, you don't deserve to be protected by the police, I call that natural selection at this point
Anonymous 01/14/25(Tue)22:59:48 No.103899436
Anonymous 01/14/25(Tue)23:00:00 No.103899441
Anonymous 01/14/25(Tue)23:01:14 No.103899451
>>103899410
but you are pleasing me with those gens lol. all I am saying is, put in the extra mile, it pays off.
but you are pleasing me with those gens lol. all I am saying is, put in the extra mile, it pays off.
Anonymous 01/14/25(Tue)23:01:24 No.103899452
Anonymous 01/14/25(Tue)23:01:26 No.103899453
>>103899441
She is abnormally tall.
She is abnormally tall.
Anonymous 01/14/25(Tue)23:02:06 No.103899459
Anonymous 01/14/25(Tue)23:04:24 No.103899481
Anonymous 01/14/25(Tue)23:07:35 No.103899506
hey guaize it's me, the real taylor swift. i was in car accident and had to do fatal plastic surgery, now i need 500 billion dollars to redeem pleastic surgery to get rid of my buttchin and return the face back to normal, please do help
Anonymous 01/14/25(Tue)23:07:39 No.103899509
>>103899451
NO, I WONT INPAINT LESS ALONE FOR A THREAD IN 4CHAN
>>103899459
I've shown my older brother the sloppiest of "AI influencers" butt-chin maxx to the extreme, and he told me "wow that looks so real, if you didn't tell me this was AI I couldn't tell"
NO, I WONT INPAINT LESS ALONE FOR A THREAD IN 4CHAN
>>103899459
I've shown my older brother the sloppiest of "AI influencers" butt-chin maxx to the extreme, and he told me "wow that looks so real, if you didn't tell me this was AI I couldn't tell"
Anonymous 01/14/25(Tue)23:08:14 No.103899513
>>103899506
top kek
top kek
Anonymous 01/14/25(Tue)23:08:46 No.103899519
Anonymous 01/14/25(Tue)23:11:31 No.103899539
>juggernaut flux
We are so starved that even plebbitnaut gets us up in a tissy
We are so starved that even plebbitnaut gets us up in a tissy
Anonymous 01/14/25(Tue)23:13:14 No.103899558
>>103899539
Ikr, I hate those periods when there's absolutely nothing to talk about, it's too calm this month, c'mon Tencent, give us the i2v model ;-;
Ikr, I hate those periods when there's absolutely nothing to talk about, it's too calm this month, c'mon Tencent, give us the i2v model ;-;
Anonymous 01/14/25(Tue)23:13:39 No.103899562
Anonymous 01/14/25(Tue)23:15:14 No.103899573
>>103899562
must be really insulting to be considered as a genuine slop AI artist when it's just your style of drawing desu
must be really insulting to be considered as a genuine slop AI artist when it's just your style of drawing desu
Anonymous 01/14/25(Tue)23:19:17 No.103899609
Anonymous 01/14/25(Tue)23:26:16 No.103899688
anyone tried to train videos on 12gb
how did it go
how did it go
Anonymous 01/14/25(Tue)23:26:23 No.103899690
Anonymous 01/14/25(Tue)23:27:40 No.103899701
Anonymous 01/14/25(Tue)23:27:49 No.103899704
Anonymous 01/14/25(Tue)23:28:14 No.103899707
>>103899562
>>103899573
I feel bad for the guy and artists in general, now AI ruins their business and many of them will have to get real jobs.
crazy how a computer can make some skills almost completely irrelevant.
>>103899573
I feel bad for the guy and artists in general, now AI ruins their business and many of them will have to get real jobs.
crazy how a computer can make some skills almost completely irrelevant.
Anonymous 01/14/25(Tue)23:28:38 No.103899711
>>103899701
Cool effect on the ground
Cool effect on the ground
Anonymous 01/14/25(Tue)23:28:58 No.103899715
Anonymous 01/14/25(Tue)23:30:08 No.103899725
>>103899711
Thanks
Thanks
Anonymous 01/14/25(Tue)23:30:18 No.103899727
>>103899688
I might do some research, see what the absolute minimum number of frames is to capture a specific motion adequately, but don't hold your breath
I might do some research, see what the absolute minimum number of frames is to capture a specific motion adequately, but don't hold your breath
Anonymous 01/14/25(Tue)23:30:24 No.103899728
Anonymous 01/14/25(Tue)23:30:32 No.103899730
Anonymous 01/14/25(Tue)23:31:32 No.103899741
Anonymous 01/14/25(Tue)23:31:52 No.103899744
Anonymous 01/14/25(Tue)23:34:17 No.103899764
>>103899707
that's why I believe we're too populated on this world, ultra automation would work fine if we were like less than a billion
that's why I believe we're too populated on this world, ultra automation would work fine if we were like less than a billion
Anonymous 01/14/25(Tue)23:35:37 No.103899773
Anonymous 01/14/25(Tue)23:36:57 No.103899783
according to Flux, this is the most generic picture an AI can generate
Anonymous 01/14/25(Tue)23:39:31 No.103899802
Anonymous 01/14/25(Tue)23:41:01 No.103899815
Anonymous 01/14/25(Tue)23:43:24 No.103899827
>>103899815
why would I pay some artist cuck when I can generate anything I want?
why would I pay some artist cuck when I can generate anything I want?
Anonymous 01/14/25(Tue)23:43:28 No.103899828
Anonymous 01/14/25(Tue)23:43:47 No.103899832
>>103899815
you don't have to be a genius to guess that if people can make AI pictures easily by themselves, they don't need real artists anymore
you don't have to be a genius to guess that if people can make AI pictures easily by themselves, they don't need real artists anymore
Anonymous 01/14/25(Tue)23:44:48 No.103899838
Anonymous 01/14/25(Tue)23:54:52 No.103899918
Hun img2vid soon, extremely soon. GET HYPED
Anonymous 01/14/25(Tue)23:55:45 No.103899925
Anonymous 01/14/25(Tue)23:55:45 No.103899926
Anonymous 01/14/25(Tue)23:56:31 No.103899930
hunyuan text2image 2.... soon..... too..... right?
Anonymous 01/14/25(Tue)23:56:51 No.103899934
>>103899828
DAMN BRO THIS BE KINO!!!!!
DAMN BRO THIS BE KINO!!!!!
Anonymous 01/14/25(Tue)23:57:51 No.103899939
A hunnyoowan just flew over my house
Anonymous 01/14/25(Tue)23:58:38 No.103899944
>>103899930
HunyuanDiT2, fully uncensored, has all the artists and celebrities, has dalle3 prompt understanding level, is better than flux and is only a 2b model, trust the plan
HunyuanDiT2, fully uncensored, has all the artists and celebrities, has dalle3 prompt understanding level, is better than flux and is only a 2b model, trust the plan
Anonymous 01/14/25(Tue)23:59:04 No.103899948
Anonymous 01/14/25(Tue)23:59:57 No.103899955
Anonymous 01/15/25(Wed)00:06:04 No.103899995
Anonymous 01/15/25(Wed)00:06:52 No.103900001
>>103899995
why? like I said no one is gonna buy anything from any artists, whether it's normal artists or AI artists, we can do thoses drawings by ourselves for free now
why? like I said no one is gonna buy anything from any artists, whether it's normal artists or AI artists, we can do thoses drawings by ourselves for free now
Anonymous 01/15/25(Wed)00:07:08 No.103900004
Anonymous 01/15/25(Wed)00:11:40 No.103900038
Anonymous 01/15/25(Wed)00:14:16 No.103900054
Anonymous 01/15/25(Wed)00:14:42 No.103900060
Anonymous 01/15/25(Wed)00:16:46 No.103900077
Anonymous 01/15/25(Wed)00:17:50 No.103900087
Anyone know where clip skip is in A1111?
Anonymous 01/15/25(Wed)00:18:33 No.103900095
Anonymous 01/15/25(Wed)00:18:42 No.103900098
>>103900077
why did she eat the bottom of the cone? is she stupid? its all gonna fall out the bottom now. she must be a stupid generation
why did she eat the bottom of the cone? is she stupid? its all gonna fall out the bottom now. she must be a stupid generation
Anonymous 01/15/25(Wed)00:47:31 No.103900247
>>103899934
(kino:1.2)
(kino:1.2)
Anonymous 01/15/25(Wed)00:52:14 No.103900271
Anonymous 01/15/25(Wed)01:06:47 No.103900354