/ldg/ - Local Diffusion General
Anonymous 01/21/25(Tue)17:22:23 | 316 comments | 98 images | 🔒 Locked
What Model is That? Edition
Discussion of Free and Open-Source Diffusion models.
Last Bread: >>103969385
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
Discussion of Free and Open-Source Diffusion models.
Last Bread: >>103969385
Newfrien Guide: https://rentry.org/ldg-lazy-getting
>UI
Metastable: https://metastable.studio
SwarmUI: https://github.com/mcmonkeyprojects
Forge: https://github.com/lllyasviel/stabl
reForge: https://github.com/Panchovix/stable
ComfyUI: https://github.com/comfyanonymous/C
InvokeAI: https://github.com/invoke-ai/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/ComfyUI-Hu
Windows: https://rentry.org/crhcqq54
Training: https://github.com/kohya-ss/musubi-
>Flux
Forge: https://github.com/lllyasviel/stabl
ComfyUI: https://comfyanonymous.github.io/Co
DeDistilled: https://huggingface.co/TheYuriLover
Pixelwave: https://civitai.com/models/141592/p
>Misc
Share Metadata: https://catbox.moe | https://litterbox.catbox.moe/
Img2Prompt: https://huggingface.co/spaces/fancy
Archive: https://rentry.org/sdg-link
Samplers: https://stable-diffusion-art.com/sa
Digital Art Software: https://krita.org/en/
Txt2Img Plugin: https://kritaaidiffusion.com/
Collagebaker: https://www.befunky.com/create/coll
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 01/21/25(Tue)17:27:41 No.103985290
Blessed thread of frenship
Anonymous 01/21/25(Tue)17:28:39 No.103985309
Anonymous 01/21/25(Tue)17:29:48 No.103985320
Imagine an anime model with Flux level comprehension
Anonymous 01/21/25(Tue)17:30:13 No.103985325
>>103985320
two more weeks
two more weeks
Anonymous 01/21/25(Tue)17:31:03 No.103985336
>>103985320
flux's japanese drawings look okay to my eyes
flux's japanese drawings look okay to my eyes
Anonymous 01/21/25(Tue)17:31:13 No.103985339
is there a 3D model from imagen anime model?
Anonymous 01/21/25(Tue)17:31:30 No.103985347
>>103985320
Flux is rigid as fuck. I'm just using it to inpaint text.
Flux is rigid as fuck. I'm just using it to inpaint text.
Anonymous 01/21/25(Tue)17:31:37 No.103985352
>>103985339
yeah
yeah
Anonymous 01/21/25(Tue)17:32:15 No.103985369
>>103985352
link?
link?
Anonymous 01/21/25(Tue)17:34:58 No.103985408
what's that thing called where women put green slime on their face. "green mask" is just a green mask
Anonymous 01/21/25(Tue)17:37:31 No.103985440
>>103985347
Imagine if it wasn't rigid tho
Imagine if it wasn't rigid tho
Anonymous 01/21/25(Tue)17:46:26 No.103985563
Anonymous 01/21/25(Tue)18:12:42 No.103985879
aaaCHOO!
Anonymous 01/21/25(Tue)18:23:08 No.103985994
Where did everyone go?
Anonymous 01/21/25(Tue)18:39:50 No.103986194
>>103985994
death march
death march
Anonymous 01/21/25(Tue)18:40:40 No.103986200
>>103986194
Please post on /sdg/ again? Schizo anon you can't post on two threads at the same time.
Please post on /sdg/ again? Schizo anon you can't post on two threads at the same time.
Anonymous 01/21/25(Tue)18:43:49 No.103986226
>>103986200
whats that?
whats that?
Anonymous 01/21/25(Tue)18:45:22 No.103986236
>>103986226
I thought you would explain.
I thought you would explain.
Anonymous 01/21/25(Tue)18:47:19 No.103986248
>>103986236
why? idk what youre on about anonie
why? idk what youre on about anonie
Anonymous 01/21/25(Tue)18:51:44 No.103986310
>>103985336
It has only a single anime style though.
It has only a single anime style though.
Anonymous 01/21/25(Tue)18:55:47 No.103986349
ani is posting updates on /sdg/ now so I'm out.
Anonymous 01/21/25(Tue)18:56:19 No.103986355
>24+ hour thread
We are back
We are back
Anonymous 01/21/25(Tue)19:00:49 No.103986409
>>103986355
Have you ever tried to post a gen? After all this is a thread about AI gens.
Have you ever tried to post a gen? After all this is a thread about AI gens.
Anonymous 01/21/25(Tue)19:01:31 No.103986422
>>103985408
skincare facemask
skincare facemask
Anonymous 01/21/25(Tue)19:02:51 No.103986437
Anonymous 01/21/25(Tue)19:03:35 No.103986443
Anonymous 01/21/25(Tue)19:12:47 No.103986552
>>103985879
bless you
bless you
Anonymous 01/21/25(Tue)19:25:03 No.103986675
>>102025928
Did this use Kelly Wells as a model?
Did this use Kelly Wells as a model?
Anonymous 01/21/25(Tue)19:39:46 No.103986825
Anonymous 01/21/25(Tue)19:40:54 No.103986837
Anonymous 01/21/25(Tue)19:41:15 No.103986841
>>103986355
as it should be. the space is in a terrible place right now, anons been ridden hard and put away wet.
as it should be. the space is in a terrible place right now, anons been ridden hard and put away wet.
Anonymous 01/21/25(Tue)19:52:08 No.103986985
>I missed the gigachad Jesus collage
f-fuck..
f-fuck..
Anonymous 01/21/25(Tue)20:12:15 No.103987195
>>103986841
And what happened to your images exactly? Let me guess you have never posted any images.
And what happened to your images exactly? Let me guess you have never posted any images.
Anonymous 01/21/25(Tue)20:15:46 No.103987232
Anonymous 01/21/25(Tue)20:17:24 No.103987249
Sorry can someone explain the difference between lora model strength and clip strength again?
Anonymous 01/21/25(Tue)20:17:34 No.103987251
>>103986985
You are Koff.
You are Koff.
Anonymous 01/21/25(Tue)20:19:19 No.103987272
>>103987249
Lora multiplier means how strong it will affect the clip conditioning, clip strength I don't know. Maybe you can multiply them and do something.
Clip -> conditioning -> result
So that's what is what.
Lora multiplier means how strong it will affect the clip conditioning, clip strength I don't know. Maybe you can multiply them and do something.
Clip -> conditioning -> result
So that's what is what.
Anonymous 01/21/25(Tue)20:20:50 No.103987284
>>103987251
is that a racial slur?
is that a racial slur?
Anonymous 01/21/25(Tue)20:21:15 No.103987287
>>103987284
What do you mean?
What do you mean?
Anonymous 01/21/25(Tue)20:23:15 No.103987312
>>103987249
model strength is more for composition or characters clip is more for styles
model strength is more for composition or characters clip is more for styles
Anonymous 01/21/25(Tue)20:23:32 No.103987315
>>103987287
i dont know what a koff is
i dont know what a koff is
Anonymous 01/21/25(Tue)20:24:19 No.103987325
>>103987315
I don't know either.
I don't know either.
Anonymous 01/21/25(Tue)20:24:31 No.103987327
Anonymous 01/21/25(Tue)20:26:08 No.103987343
Anonymous 01/21/25(Tue)20:31:32 No.103987394
tfw waiting for new model
Anonymous 01/21/25(Tue)20:32:27 No.103987401
>>103987394
bruh I've been waiting for a new UI forever
bruh I've been waiting for a new UI forever
Anonymous 01/21/25(Tue)20:33:12 No.103987416
>>103987325
ok then
ok then
Anonymous 01/21/25(Tue)20:36:49 No.103987471
>>103987416
very nice upscale, smooth
very nice upscale, smooth
Anonymous 01/21/25(Tue)20:38:57 No.103987500
>>103987401
just write your own inference script brah
just write your own inference script brah
Anonymous 01/21/25(Tue)20:39:14 No.103987504
>>103987416
I know you...
I know you...
Anonymous 01/21/25(Tue)20:41:23 No.103987527
>>103987416
I don't know you.
I don't know you.
Anonymous 01/21/25(Tue)20:41:52 No.103987536
Calm down anon
Anonymous 01/21/25(Tue)20:42:22 No.103987540
>>103987471
upscaling is so unpredictable, i really don't know how it gets so clean sometimes.
>>103987504
You don't know you, you only know yourself.
>>103987527
I don't know me.
upscaling is so unpredictable, i really don't know how it gets so clean sometimes.
>>103987504
You don't know you, you only know yourself.
>>103987527
I don't know me.
Anonymous 01/21/25(Tue)20:43:04 No.103987552
>>103987540
You are catjak....
You are catjak....
Anonymous 01/21/25(Tue)20:45:14 No.103987568
>>103987540
>upscaling is so unpredictable, i really don't know how it gets so clean sometimes.
you have found good settings
>upscaling is so unpredictable, i really don't know how it gets so clean sometimes.
you have found good settings
Anonymous 01/21/25(Tue)20:45:17 No.103987570
>>103987540
Ran...
Ran...
Anonymous 01/21/25(Tue)20:46:09 No.103987583
>>103985188
Gigachadjesus based
Gigachadjesus based
Anonymous 01/21/25(Tue)20:48:29 No.103987600
Anonymous 01/21/25(Tue)20:49:57 No.103987614
>>103987600
Maybe you didn't hear that A1111 has been depreceated since one year ago.
Maybe you didn't hear that A1111 has been depreceated since one year ago.
Anonymous 01/21/25(Tue)20:52:33 No.103987635
>>103987600
Thanks gotta check it out.
I upscale these >>103987327 with i2i (controlnet tile). Haven't used hiresfix for a while.
Thanks gotta check it out.
I upscale these >>103987327 with i2i (controlnet tile). Haven't used hiresfix for a while.
Anonymous 01/21/25(Tue)20:53:30 No.103987651
>>103987635
Thank God you are from the US. It's your right to be angry.
Thank God you are from the US. It's your right to be angry.
Anonymous 01/21/25(Tue)20:55:42 No.103987667
Anonymous 01/21/25(Tue)20:55:55 No.103987671
>>103987635
yeah i dunno man forge's hiresfix has always just werked, got memed into comfyui for a while but i went back to old reliable.
yeah i dunno man forge's hiresfix has always just werked, got memed into comfyui for a while but i went back to old reliable.
Anonymous 01/21/25(Tue)20:57:12 No.103987687
>>103987671
Catjak...
Catjak...
Anonymous 01/21/25(Tue)20:59:18 No.103987708
>>103987667
If you want to succeed - obtain 4090 and spam 1girls and anime.
If you want to succeed - obtain 4090 and spam 1girls and anime.
Anonymous 01/21/25(Tue)20:59:41 No.103987712
>>103987667
1. see anon img you like
2. ask anon for catbox
3. load catbox into UI and press generate
if 2 doesnt work then... shit... lurk moar i guess
1. see anon img you like
2. ask anon for catbox
3. load catbox into UI and press generate
if 2 doesnt work then... shit... lurk moar i guess
Anonymous 01/21/25(Tue)21:16:23 No.103987887
>>103987540
best thing i ever did was add anytest controlnet to hiresfix workflow
best thing i ever did was add anytest controlnet to hiresfix workflow
Anonymous 01/21/25(Tue)21:20:50 No.103987919
>>103987887
how does that work? It's these right? https://huggingface.co/2vXpSwA7/iroiro-lora/tree/main
how does that work? It's these right? https://huggingface.co/2vXpSwA7/iro
Anonymous 01/21/25(Tue)21:22:15 No.103987930
>>103987600
https://openmodeldb.info/models/4x-IllustrationJaNai-V1-DAT2 replaced 4x-Ultrasharp for me personally
https://openmodeldb.info/models/4x-
Anonymous 01/21/25(Tue)21:25:16 No.103987961
>>103987887
>>103987919
interesting, how does that work/whats it do?
>>103987930
that looks really nice, gonna try that now,ultrasharp is old as fuck from what i remember.
>>103987919
interesting, how does that work/whats it do?
>>103987930
that looks really nice, gonna try that now,ultrasharp is old as fuck from what i remember.
Anonymous 01/21/25(Tue)21:26:07 No.103987972
>>103987919
yeah i use one of the anytest v4s, 0.9 power, end_percent 0.89, using the base gen as the control image
hiresfix is just a script for img2img with various settings, upscale w/ upscale model, scale that to final resolution, then do a partial denoise on that to try and get the model to connect some dots. you can insert a controlnet into that process but maybe the extra clicks aren't worth it on your program
>>103987930
2x-animesharpv3 for life
yeah i use one of the anytest v4s, 0.9 power, end_percent 0.89, using the base gen as the control image
hiresfix is just a script for img2img with various settings, upscale w/ upscale model, scale that to final resolution, then do a partial denoise on that to try and get the model to connect some dots. you can insert a controlnet into that process but maybe the extra clicks aren't worth it on your program
>>103987930
2x-animesharpv3 for life
Anonymous 01/21/25(Tue)21:27:33 No.103987990
Anonymous 01/21/25(Tue)21:28:08 No.103987999
>ESRGAN
Boomer tech
Boomer tech
Anonymous 01/21/25(Tue)21:28:55 No.103988006
>>103987990
it's not compatible with my version of comfy and i dont feel like updating
it's not compatible with my version of comfy and i dont feel like updating
Anonymous 01/21/25(Tue)21:37:47 No.103988090
>>103987972
>anytest v4s
this? https://huggingface.co/2vXpSwA7/iroiro-lora/blob/main/test_controlnet2/CN-anytest_v4-marged_am_dim256.safetensors
Is it like tile, style copy or what? Empty model card like usual
>anytest v4s
this? https://huggingface.co/2vXpSwA7/iro
Is it like tile, style copy or what? Empty model card like usual
Anonymous 01/21/25(Tue)21:39:42 No.103988115
>>103988006
I just pulled and it still doesn't work. :(
I just pulled and it still doesn't work. :(
Anonymous 01/21/25(Tue)21:40:00 No.103988120
Holy fuck I'm so dumb I wasn't using the face detailer.
Anonymous 01/21/25(Tue)21:41:06 No.103988133
>>103988120
Many such cases!
Many such cases!
Anonymous 01/21/25(Tue)21:41:11 No.103988134
>>103988090
i use the fat 2.5GB one, idk about the distilled versions
you throw it any image and it controls the output towards that image, no preprocessing
for hiresfix cntext it effectively stops major mutations
i use the fat 2.5GB one, idk about the distilled versions
you throw it any image and it controls the output towards that image, no preprocessing
for hiresfix cntext it effectively stops major mutations
Anonymous 01/21/25(Tue)21:42:14 No.103988149
RealESRGAN_x4Plus Anime 6B is still the best general upscaler right?
Anonymous 01/21/25(Tue)21:42:59 No.103988152
>>103985994
they're all in lmg playing with deepseek r1
they're all in lmg playing with deepseek r1
Anonymous 01/21/25(Tue)21:44:47 No.103988166
>>103988115
You would need to reinstall dependencies in pip or upgrade Spandrel manually because the support comes from a library instead of inside ComfyUI itself. Do a "pip install -U spandrel" in your venv or Python install.
You would need to reinstall dependencies in pip or upgrade Spandrel manually because the support comes from a library instead of inside ComfyUI itself. Do a "pip install -U spandrel" in your venv or Python install.
Anonymous 01/21/25(Tue)21:46:26 No.103988183
Didn't Cosmos just basically kill Flux now that you can fine-tune it to make images?
Anonymous 01/21/25(Tue)21:47:17 No.103988191
>>103988183
show me a good cosmos gen that rivals flux
show me a good cosmos gen that rivals flux
Anonymous 01/21/25(Tue)21:48:57 No.103988206
Not sure why but every time my masked image runs through ultimate SD upscale it ends up affecting details of the whole image instead of just the masked area.
Anonymous 01/21/25(Tue)22:00:25 No.103988319
>>103988166
Thanks.
https://imgsli.com/MzQwMTMx/3/2
I did a comparison between >>103987930 and >>103987990 using a downscaled gen from this thread. Since JaNai is 4x while AnimeSharp is 2x I also included a double pass of the latter. Unless I did something wrong or am using the wrong version, it doesn't smooth out the pixelization on the second run however.
Thanks.
https://imgsli.com/MzQwMTMx/3/2
I did a comparison between >>103987930 and >>103987990 using a downscaled gen from this thread. Since JaNai is 4x while AnimeSharp is 2x I also included a double pass of the latter. Unless I did something wrong or am using the wrong version, it doesn't smooth out the pixelization on the second run however.
Anonymous 01/21/25(Tue)22:02:55 No.103988334
>>103988206
What do you mean?
What do you mean?
Anonymous 01/21/25(Tue)22:05:24 No.103988358
>>103988149
4x-UltraSharp.pth
4x-UltraSharp.pth
Anonymous 01/21/25(Tue)22:07:49 No.103988377
trending on attestation , by Greg Rutkowski
Anonymous 01/21/25(Tue)22:08:23 No.103988384
seems like janai_v1_esrgan_135k is really good at preserving certain details while adding to it, doesn't royally fuck with faces or anything like ive noticed with ultrasharp sometimes
i wonder if its less/more sensitive to denoising from that..
i wonder if its less/more sensitive to denoising from that..
Anonymous 01/21/25(Tue)22:09:22 No.103988394
Anonymous 01/21/25(Tue)22:09:45 No.103988396
>>103988319
It's not supposed to. It does no restoration whatsoever, per its description.
>Description: This is a successor to AnimeSharpV3 based on RCAN instead of ESRGAN. It outperforms both versions of AnimeSharpV3 in every capacity. It's sharper, retains even more detail, and has very few artifacts. It is extremely faithful to the input image, even with heavily compressed inputs.
It's not supposed to. It does no restoration whatsoever, per its description.
>Description: This is a successor to AnimeSharpV3 based on RCAN instead of ESRGAN. It outperforms both versions of AnimeSharpV3 in every capacity. It's sharper, retains even more detail, and has very few artifacts. It is extremely faithful to the input image, even with heavily compressed inputs.
Anonymous 01/21/25(Tue)22:11:30 No.103988407
Anonymous 01/21/25(Tue)22:18:16 No.103988458
>>103988334
I took the image on the left and inpainted over the the left girl to turn her brown but when it's run through the upscaler the prompt I used for her face affected both girls.
I took the image on the left and inpainted over the the left girl to turn her brown but when it's run through the upscaler the prompt I used for her face affected both girls.
Anonymous 01/21/25(Tue)22:19:40 No.103988470
>>103988458
I any case just edit the image in Photoshop.
I any case just edit the image in Photoshop.
Anonymous 01/21/25(Tue)22:20:13 No.103988479
>>103988470
That's typically what I've been doing but it's still a pain.
That's typically what I've been doing but it's still a pain.
Anonymous 01/21/25(Tue)22:21:40 No.103988490
>>103988458
You can img2img and do a little photoshop repair in between. You can use mouse and rubber stamp brush with a mask, and it's fine for most of these anime trash images.
You can img2img and do a little photoshop repair in between. You can use mouse and rubber stamp brush with a mask, and it's fine for most of these anime trash images.
Anonymous 01/21/25(Tue)22:22:30 No.103988497
>>103988458
>but when it's run through the upscaler the prompt I used for her face affected both girls.
are you using a face detailer? maybe a connection got crossed
>but when it's run through the upscaler the prompt I used for her face affected both girls.
are you using a face detailer? maybe a connection got crossed
Anonymous 01/21/25(Tue)22:22:50 No.103988501
>>103988479
That's what you need to pay for any proper image. More you work, more you need to realize it's not just a click.
That's what you need to pay for any proper image. More you work, more you need to realize it's not just a click.
Anonymous 01/21/25(Tue)22:24:07 No.103988513
>>103988497
It happens before the face detailer. So even if I turn it off it still happens.
>>103988490
Usually I layer the before and after images on top of each other and erase around the area I want to keep and just merge them.
It happens before the face detailer. So even if I turn it off it still happens.
>>103988490
Usually I layer the before and after images on top of each other and erase around the area I want to keep and just merge them.
Anonymous 01/21/25(Tue)22:27:40 No.103988542
>>103988513
This depends. I am used to work with comp softwares and they are all progressive.
With Photoshop, it doesn't matter what you do but if you do and deliver it. This means you can be dirty with the underlying layers etc. because it's not procedural workflow.
This depends. I am used to work with comp softwares and they are all progressive.
With Photoshop, it doesn't matter what you do but if you do and deliver it. This means you can be dirty with the underlying layers etc. because it's not procedural workflow.
Anonymous 01/21/25(Tue)22:28:42 No.103988554
>>103988513
I mean you work on it and layer as good as you think it is.
I mean you work on it and layer as good as you think it is.
Anonymous 01/21/25(Tue)22:29:34 No.103988567
>>103988458
there's always going to be some prompt bleeding in more complex images when upscaling, so i recommend you do major changes at base gen and detailing/fixes after the upscale
there's always going to be some prompt bleeding in more complex images when upscaling, so i recommend you do major changes at base gen and detailing/fixes after the upscale
Anonymous 01/21/25(Tue)22:32:50 No.103988604
>>103988513
Can you post an image without any edits?
Can you post an image without any edits?
Anonymous 01/21/25(Tue)22:34:36 No.103988618
Anonymous 01/21/25(Tue)22:35:27 No.103988628
>>103988618
Thank you!
Thank you!
Anonymous 01/21/25(Tue)22:40:31 No.103988676
Anonymous 01/21/25(Tue)22:44:44 No.103988714
Anonymous 01/21/25(Tue)22:46:27 No.103988729
Anonymous 01/21/25(Tue)22:58:44 No.103988849
Anonymous 01/21/25(Tue)23:00:52 No.103988870
I used to know so much about this shit but now I took a year off and I feel like a monster, monster retard
Anonymous 01/21/25(Tue)23:02:22 No.103988886
>>103988870
One year is quite a long time here
One year is quite a long time here
Anonymous 01/21/25(Tue)23:04:20 No.103988911
>>103988886
Man one month is a long time, I remember when I was actively genning if I took a week or two off I feel like I missed a lot. Can I get a rundown, what's good, what models are people using these days? What new custom nodes do I need to have to live?
Man one month is a long time, I remember when I was actively genning if I took a week or two off I feel like I missed a lot. Can I get a rundown, what's good, what models are people using these days? What new custom nodes do I need to have to live?
Anonymous 01/21/25(Tue)23:19:56 No.103989059
Anonymous 01/21/25(Tue)23:31:24 No.103989150
>>103988396
Would've helped me to read that, huh.
Would've helped me to read that, huh.
Anonymous 01/21/25(Tue)23:32:45 No.103989167
Anonymous 01/21/25(Tue)23:35:17 No.103989196
Anonymous 01/21/25(Tue)23:35:52 No.103989203
Need more gigachad Jesus.
Anonymous 01/21/25(Tue)23:39:01 No.103989232
>>103989203
Sorry, all I've got is dorf
Sorry, all I've got is dorf
Anonymous 01/21/25(Tue)23:53:21 No.103989359
Anonymous 01/22/25(Wed)00:20:00 No.103989567
Anonymous 01/22/25(Wed)00:32:06 No.103989642
Anonymous 01/22/25(Wed)00:42:33 No.103989708
All my gens with this model come up with several artifacts, that means there is something fucked up with the training right? no matter what sampler or scheduler I use, it looks like this
Anonymous 01/22/25(Wed)00:44:40 No.103989721
Anonymous 01/22/25(Wed)00:45:34 No.103989728
Anonymous 01/22/25(Wed)00:49:45 No.103989756
Anonymous 01/22/25(Wed)00:52:13 No.103989774
What does rescale CFG mean? In the CADs extentsion(though I think I saw it in another extension).
I'm stupid so I don't understand. I interprit as "uncheck this to increase schizophrenia score by 999".
I'm stupid so I don't understand. I interprit as "uncheck this to increase schizophrenia score by 999".
Anonymous 01/22/25(Wed)00:53:39 No.103989783
Anonymous 01/22/25(Wed)00:54:10 No.103989788
Anonymous 01/22/25(Wed)00:54:56 No.103989795
>>103989708
catbox? to answer your question it could just be the model. I remember some pony realism model always giving me weird splotches, you can see one in this clip. I think it was cfg related.
catbox? to answer your question it could just be the model. I remember some pony realism model always giving me weird splotches, you can see one in this clip. I think it was cfg related.
Anonymous 01/22/25(Wed)00:58:05 No.103989805
Anonymous 01/22/25(Wed)00:59:53 No.103989821
Anonymous 01/22/25(Wed)01:02:32 No.103989837
>>103989795
I remember the ponyrealism model giving out the same artifacts
I remember the ponyrealism model giving out the same artifacts
Anonymous 01/22/25(Wed)01:04:27 No.103989851
Anonymous 01/22/25(Wed)01:05:53 No.103989864
Anonymous 01/22/25(Wed)01:08:38 No.103989892
>>103989837
>that was the model's name
kek yeah I think it was. I got this prompt from an anon and I'm pretty sure lowering the cfg fixed it but it had some adverse effect I didn't like, so I just embraced the artifacts.
And please give a catbox anon.
>that was the model's name
kek yeah I think it was. I got this prompt from an anon and I'm pretty sure lowering the cfg fixed it but it had some adverse effect I didn't like, so I just embraced the artifacts.
And please give a catbox anon.
Anonymous 01/22/25(Wed)01:19:25 No.103989962
![1737346159847925[1]](https://i.4cdn.org/g/1737526765037567s.jpg)
Anonymous 01/22/25(Wed)01:24:21 No.103990001
Anonymous 01/22/25(Wed)01:27:35 No.103990032
Anonymous 01/22/25(Wed)01:32:10 No.103990068
Anonymous 01/22/25(Wed)01:54:03 No.103990222
Which is the best image to image model?
Anonymous 01/22/25(Wed)02:14:10 No.103990369
>>103989708
It could be the training of the model or some lora you use, but it is always really hard to tell.
It could be the training of the model or some lora you use, but it is always really hard to tell.
Anonymous 01/22/25(Wed)02:32:48 No.103990465
Total Coomer Death
Anonymous 01/22/25(Wed)03:15:16 No.103990780
>>103988911
check OP
check OP
Anonymous 01/22/25(Wed)03:53:55 No.103991061
Anonymous 01/22/25(Wed)04:10:59 No.103991198
Testing out Hunyuan. This video took 367 seconds. Is that reasonable, or really bad?
Is there a good workflow for generating Hunyuan videos much faster, and with only 12 GB of VRAM?
(already blockswapping, already using sageattention, can't use torchcompile because I get an error message that I can't solve.)
Is there a good workflow for generating Hunyuan videos much faster, and with only 12 GB of VRAM?
(already blockswapping, already using sageattention, can't use torchcompile because I get an error message that I can't solve.)
Anonymous 01/22/25(Wed)04:12:48 No.103991218
>>103991198
teacache could shave a minute or two off
teacache could shave a minute or two off
Anonymous 01/22/25(Wed)04:22:27 No.103991279
I've solved my issue.
Anonymous 01/22/25(Wed)05:20:43 No.103991658
Anonymous 01/22/25(Wed)05:44:20 No.103991782
>>103985188
I have only 12GB VRAM on a 3060 but I have a few distant aunts who I need to coom to
Can I do any kind of video on this?
I have only 12GB VRAM on a 3060 but I have a few distant aunts who I need to coom to
Can I do any kind of video on this?
Anonymous 01/22/25(Wed)05:49:36 No.103991815
Anonymous 01/22/25(Wed)05:51:34 No.103991827
>>103991782
yes it's good for 3 seconds of 720x560 or lower in a reasonable timeframe, gets ridiculous after that but still possible
yes it's good for 3 seconds of 720x560 or lower in a reasonable timeframe, gets ridiculous after that but still possible
Anonymous 01/22/25(Wed)05:53:03 No.103991839
>>103991827
Thanks anon! I'll go ahead
>>103991815
And I'm a thirdie who bought a 3060 for 280$ a few months ago
We all do what we gotta do
Thanks anon! I'll go ahead
>>103991815
And I'm a thirdie who bought a 3060 for 280$ a few months ago
We all do what we gotta do
Anonymous 01/22/25(Wed)06:06:24 No.103991918
I have frighteningly little sfw clips to post, but this is the upper end of what a 3060 can do before you're stuck waiting days for it to process
and this is begrudgingly using kijai's wrapper, I don't know how but it stretches resources further than the native nodes in all the tests I've done. Maybe ggufs are good for 16+vram but it's not enough to offset the vram usage of comfys nodes with only 12
and this is begrudgingly using kijai's wrapper, I don't know how but it stretches resources further than the native nodes in all the tests I've done. Maybe ggufs are good for 16+vram but it's not enough to offset the vram usage of comfys nodes with only 12
Anonymous 01/22/25(Wed)06:19:08 No.103991999
Anonymous 01/22/25(Wed)06:28:23 No.103992078
>>103991782
Yeah, I use that card, too. I make videos, but it takes around 330~360 seconds for me since I can't figure out how to work the stuff that makes it faster. I can't even figure out what "teacache" is.
Yeah, I use that card, too. I make videos, but it takes around 330~360 seconds for me since I can't figure out how to work the stuff that makes it faster. I can't even figure out what "teacache" is.
Anonymous 01/22/25(Wed)07:06:06 No.103992404
Anonymous 01/22/25(Wed)07:10:22 No.103992452
>>103991815
>>103992078
>Prompt executed in 392.88 seconds
For 768x512 image (like in single frame) on my poverty hardware.
>>103992078
>Prompt executed in 392.88 seconds
For 768x512 image (like in single frame) on my poverty hardware.
Anonymous 01/22/25(Wed)07:51:54 No.103992811
Anonymous 01/22/25(Wed)07:59:35 No.103992880
Why are there two difusion generals?
Anonymous 01/22/25(Wed)08:08:29 No.103992973
What's the <current year> best method to create consistent art style images for example like pic related? What I did before was to create a LoRA but it always worked just meh. Now that new flux seems promissing. Would it be a similar workflow?
Anonymous 01/22/25(Wed)08:10:51 No.103992994
What are the popular models now? Lora still works? Any new stuff?
Anonymous 01/22/25(Wed)08:14:49 No.103993038
>>103992880
I was around when there was only one general.
There are a few posters in that general who constantly spammed images of their OCs (purple witch, angry girl with blonde girls) until it was unbearable
There was also some drama where the thread baker kept including a link in the OP that doxed someone, and the mods refused to do anything about it
It was also dominated by coomer 1girl slop
Eventually, some anons decided to just make their own general without any of the bullshit in the other general
Sadly, this thread also has a faggot who won't stop posting his OC, people still bitch about the OP when their gen isn't included in the collage, there is plenty of 1girl slop here, too...
The two threads are basically clones now
But at least that fucking purple witch faggot isn't here
I was around when there was only one general.
There are a few posters in that general who constantly spammed images of their OCs (purple witch, angry girl with blonde girls) until it was unbearable
There was also some drama where the thread baker kept including a link in the OP that doxed someone, and the mods refused to do anything about it
It was also dominated by coomer 1girl slop
Eventually, some anons decided to just make their own general without any of the bullshit in the other general
Sadly, this thread also has a faggot who won't stop posting his OC, people still bitch about the OP when their gen isn't included in the collage, there is plenty of 1girl slop here, too...
The two threads are basically clones now
But at least that fucking purple witch faggot isn't here
Anonymous 01/22/25(Wed)08:15:32 No.103993045
>>103989642
God damn I wish real women were shaped like that
Thank you AI! You convinced me to buy an AI robo wife!
God damn I wish real women were shaped like that
Thank you AI! You convinced me to buy an AI robo wife!
Anonymous 01/22/25(Wed)08:16:31 No.103993051
>>103992973
I don't think that any current model will oneshot images like that too often. Lora, composite, controlnet will help. I'd go with NoobAI v-pred model and train lora using that as base and then iterate. Flux can work, but you likely have to train lora for it using civitai online service because it needs so much resources.
I don't think that any current model will oneshot images like that too often. Lora, composite, controlnet will help. I'd go with NoobAI v-pred model and train lora using that as base and then iterate. Flux can work, but you likely have to train lora for it using civitai online service because it needs so much resources.
Anonymous 01/22/25(Wed)08:19:23 No.103993093
>>103992994
Are you that anon who really loved posting Tachibana Kyouka? Well, not that I can blame you...
Pony was the top model for a while, but other models have risen to take its spot. I prefer using CatTower.
Are you that anon who really loved posting Tachibana Kyouka? Well, not that I can blame you...
Pony was the top model for a while, but other models have risen to take its spot. I prefer using CatTower.
Anonymous 01/22/25(Wed)08:22:29 No.103993128
>>103993093
For a short while. I made her Lora, but it looks like civitai removed it and someone else posted their own. I don't have mine anymore. I wonder if some anon saved it.
For a short while. I made her Lora, but it looks like civitai removed it and someone else posted their own. I don't have mine anymore. I wonder if some anon saved it.
Anonymous 01/22/25(Wed)08:25:21 No.103993154
>>103989642
What model did you use for this? box?
What model did you use for this? box?
Anonymous 01/22/25(Wed)08:27:30 No.103993178
Anonymous 01/22/25(Wed)08:28:41 No.103993192
>>103993128
The latest models have so much data that I rarely have to use loras anymore. I just post "zeronis, ayanami rei" and I get ayanami rei who looks like she was drawn by zeronis, without using a zeronis lora or a rei lora.
However, Tachibana Kyouka is kinda obscure, and CatTower (for example) doesn't know who she is and doesn't generate her by name alone. Might need a lora to make pictures of her.
The latest models have so much data that I rarely have to use loras anymore. I just post "zeronis, ayanami rei" and I get ayanami rei who looks like she was drawn by zeronis, without using a zeronis lora or a rei lora.
However, Tachibana Kyouka is kinda obscure, and CatTower (for example) doesn't know who she is and doesn't generate her by name alone. Might need a lora to make pictures of her.
Anonymous 01/22/25(Wed)08:30:31 No.103993220
>>103993192
I'll try some combinations. Thanks, anon.
I'll try some combinations. Thanks, anon.
Anonymous 01/22/25(Wed)08:30:59 No.103993227
>>103993154
So this is the hardcore "technical discussion" and this is why /sdg/ is such a bad thread.
>prompt?
>catbox
So this is the hardcore "technical discussion" and this is why /sdg/ is such a bad thread.
>prompt?
>catbox
Anonymous 01/22/25(Wed)08:31:24 No.103993229
Anonymous 01/22/25(Wed)08:32:11 No.103993244
>>103993220
not him but filter "illustrious" models on civitai for the new stuff
not him but filter "illustrious" models on civitai for the new stuff
Anonymous 01/22/25(Wed)08:32:13 No.103993245
>>103993038
Ran, you are the reason why /ldg/ exists. Unfortunately for you, the attention is still not enough.
Ran, you are the reason why /ldg/ exists. Unfortunately for you, the attention is still not enough.
Anonymous 01/22/25(Wed)08:32:43 No.103993254
>>103993227
welcome back
welcome back
Anonymous 01/22/25(Wed)08:35:45 No.103993288
>use karras, they said
Anonymous 01/22/25(Wed)08:36:59 No.103993304
>>103991658
thumb
thumb
Anonymous 01/22/25(Wed)08:41:30 No.103993358
>>103993254
Thanks, brother.
Thanks, brother.
Anonymous 01/22/25(Wed)08:42:00 No.103993363
>>103993229
I've heard it said that most modern models/checkpoints have a vae baked in and don't even need separate standalone vaes anymore. But I'd like someone else to corroborate that information before I believe it.
I've heard it said that most modern models/checkpoints have a vae baked in and don't even need separate standalone vaes anymore. But I'd like someone else to corroborate that information before I believe it.
Anonymous 01/22/25(Wed)08:47:53 No.103993430
>>103993363
It's true, and can be confirmed by the model size. A 6.4 GB SDXL model probably doesn't have a VAE baked in, whereas a 6.6 GB model probably does.
It's true, and can be confirmed by the model size. A 6.4 GB SDXL model probably doesn't have a VAE baked in, whereas a 6.6 GB model probably does.
Anonymous 01/22/25(Wed)08:48:24 No.103993437
>>103993363
How many images have you generated? Zero?
How many images have you generated? Zero?
Anonymous 01/22/25(Wed)08:48:48 No.103993440
>>103993288
Lookin' like some Saya no Uta shit right there
Lookin' like some Saya no Uta shit right there
Anonymous 01/22/25(Wed)09:02:45 No.103993612
>>103993244
People abandoned noobai?
anyway, can sdxl proper or whatever, illustrious.... can it do holding hands in various positions correctly, or is it hit or miss?
People abandoned noobai?
anyway, can sdxl proper or whatever, illustrious.... can it do holding hands in various positions correctly, or is it hit or miss?
Anonymous 01/22/25(Wed)09:04:33 No.103993633
>>103993437
There's some hilarious irony in mocking another poster for not generating images while not sharing anything that you've made.
There's some hilarious irony in mocking another poster for not generating images while not sharing anything that you've made.
Anonymous 01/22/25(Wed)09:11:10 No.103993680
>>103993612
I dunno I skipped noobai and don't really pose hands.
I dunno I skipped noobai and don't really pose hands.
Anonymous 01/22/25(Wed)09:19:38 No.103993749
Fuck you mom and dad I'm caching the latents
Anonymous 01/22/25(Wed)09:20:17 No.103993754
>>103993749
DO NOT CACHE THE LATENTS BLOODY BASTARD
DO NOT CACHE THE LATENTS BLOODY BASTARD
Anonymous 01/22/25(Wed)09:35:49 No.103993918
Anonymous 01/22/25(Wed)09:36:48 No.103993932
>>103993612
>>103993680
>why a thing exists
>i don't know i just wanted to reply
The hardcore discussion of Generative Arts, this is the core /ldg/ experience as it seems.
>>103993680
>why a thing exists
>i don't know i just wanted to reply
The hardcore discussion of Generative Arts, this is the core /ldg/ experience as it seems.
Anonymous 01/22/25(Wed)09:38:39 No.103993953
>>103993932
it's almost like you can't trust most tards here to not act like autists
but that's the experience and its fun
it's almost like you can't trust most tards here to not act like autists
but that's the experience and its fun
Anonymous 01/22/25(Wed)09:43:15 No.103993987
>>103993953
pedophile
pedophile
Anonymous 01/22/25(Wed)09:43:52 No.103993995
>>103993932
Can you gen hand holding with any model, reliably, with various positions?
Can you gen hand holding with any model, reliably, with various positions?
Anonymous 01/22/25(Wed)09:47:06 No.103994027
>>103993987
angels have been around longer than what is considered underage
angels have been around longer than what is considered underage
Anonymous 01/22/25(Wed)09:48:36 No.103994040
>>103993987
Thanks boss
Thanks boss
Anonymous 01/22/25(Wed)09:53:20 No.103994087
>>103994027
drunk again and this early?
drunk again and this early?
Anonymous 01/22/25(Wed)10:02:23 No.103994171
it's going great...
Anonymous 01/22/25(Wed)10:05:28 No.103994207
>>103994171
what's your cfg at
what's your cfg at
Anonymous 01/22/25(Wed)10:06:28 No.103994223
>>103994207
1-2
1-2
Anonymous 01/22/25(Wed)10:06:53 No.103994234
>>103994223
sorry, meant the denoising strength, what's that value at
sorry, meant the denoising strength, what's that value at
Anonymous 01/22/25(Wed)10:08:21 No.103994253
Does GPU "performance" make a differnce in local models or is it mainly vram amount.
For example will a 5080 perform better/faster than a 4080 despite the same vram amount?
For example will a 5080 perform better/faster than a 4080 despite the same vram amount?
Anonymous 01/22/25(Wed)10:10:26 No.103994282
Anonymous 01/22/25(Wed)10:10:45 No.103994284
>>103994253
Yeah. For example 4070 ti super is around same speed or faster with sdxl based models compared to 3090.
Yeah. For example 4070 ti super is around same speed or faster with sdxl based models compared to 3090.
Anonymous 01/22/25(Wed)10:12:45 No.103994313
>>103994284
Is GTX 1650 any good? Using memory compression doubles its vram to 8GB.
Is GTX 1650 any good? Using memory compression doubles its vram to 8GB.
Anonymous 01/22/25(Wed)10:14:36 No.103994337
flux prompting has me utterly mystified.
Anonymous 01/22/25(Wed)10:15:08 No.103994341
>>103994313
I'd say 3060 12gb is the minimum
I'd say 3060 12gb is the minimum
Anonymous 01/22/25(Wed)10:16:08 No.103994350
ack
Anonymous 01/22/25(Wed)10:19:57 No.103994399
>>103994313
kek if you're really throwing out the idea of a 1650 just buy a 1080ti and call it a day
3060 12gb for $399 is NOT worth it
kek if you're really throwing out the idea of a 1650 just buy a 1080ti and call it a day
3060 12gb for $399 is NOT worth it
Anonymous 01/22/25(Wed)10:21:45 No.103994418
Anonymous 01/22/25(Wed)10:22:01 No.103994423
>>103994282
lower your denoising to like .3
lower your denoising to like .3
Anonymous 01/22/25(Wed)10:22:29 No.103994432
>>103994399
*300, saw a few sell for 399 but average price is floating in the 300s which is nutso
i think the used market is going into hyperdrive now with the absolute state of things, just checked into the 1080 ti prices and they're floating between cheaper than last i checked, and some Semitic individuals selling for $200~
*300, saw a few sell for 399 but average price is floating in the 300s which is nutso
i think the used market is going into hyperdrive now with the absolute state of things, just checked into the 1080 ti prices and they're floating between cheaper than last i checked, and some Semitic individuals selling for $200~
Anonymous 01/22/25(Wed)10:23:40 No.103994446
>a provocative illustration of a man and a woman standing together, holding hands, and barefoot. The woman wears a black bra, a flowing black skirt, and has long, brown hair that cascades down her back. Her bright, brown eyes sparkle with desire, and her skin has a warm, sun-kissed glow with noticeable tanlines. The man is dressed in a crisp, white shirt with a vibrant red bowtie, and his white pupils seem to gleam with intensity. The woman is bound and restrained with chains, her mouth open in a suggestive expression, conveying a sense of surrender and intimacy. Capture the details of their bare feet, including the woman's toes and soles, and the man's feet, as they stand together in a pose that exudes passion and connection, with a clear emphasis on BDSM and restraint.
>>103994223
You're not supposed to know.
>>103994223
You're not supposed to know.
Anonymous 01/22/25(Wed)10:28:19 No.103994512
>>103994423
It won't matter because that's an initial gen you moron.
It won't matter because that's an initial gen you moron.
Anonymous 01/22/25(Wed)10:29:16 No.103994526
>>103994512
Are you new to this?
Are you new to this?
Anonymous 01/22/25(Wed)10:30:02 No.103994535
>>103994418
Catbox?
Catbox?
Anonymous 01/22/25(Wed)10:30:38 No.103994545
>>103994446
You can get the same result with literally 5 tags. Retards think they need to larp a failed novelist in order to generate bad AI images.
You can get the same result with literally 5 tags. Retards think they need to larp a failed novelist in order to generate bad AI images.
Anonymous 01/22/25(Wed)10:31:39 No.103994557
>>103994526
Why are you asking?
Why are you asking?
Anonymous 01/22/25(Wed)10:31:58 No.103994561
>>103994423
It's just noise.
It's just noise.
Anonymous 01/22/25(Wed)10:32:59 No.103994574
>>103994545
I used ai to turn danbooru tags into a sentence, it's garbled either way. why?
I used ai to turn danbooru tags into a sentence, it's garbled either way. why?
Anonymous 01/22/25(Wed)10:36:17 No.103994610
https://files.catbox.moe/tzcew8.png
Anonymous 01/22/25(Wed)10:37:20 No.103994618
https://files.catbox.moe/tu3ctf.png
Anonymous 01/22/25(Wed)10:38:28 No.103994631
https://files.catbox.moe/xkxc7s.png
Anonymous 01/22/25(Wed)10:38:46 No.103994636
>>103994618
Very deboesque gen.
Very deboesque gen.
Anonymous 01/22/25(Wed)10:39:47 No.103994646
>>103994618
Can you get hand holding to work right?
Can you get hand holding to work right?
Anonymous 01/22/25(Wed)10:40:58 No.103994658
https://files.catbox.moe/jrghka.png
>>103994636
I do not know what a "deboesque gen" means, but I'll assume it's a compliment. Thanks fren!
>>103994646
I'm at work right now but I'll try hand-holding later on
>>103994636
I do not know what a "deboesque gen" means, but I'll assume it's a compliment. Thanks fren!
>>103994646
I'm at work right now but I'll try hand-holding later on
Anonymous 01/22/25(Wed)10:44:01 No.103994695
>>103994658
Anon wants his hand to be held.
Anon wants his hand to be held.
Anonymous 01/22/25(Wed)10:44:01 No.103994696
https://files.catbox.moe/515owx.png
Anonymous 01/22/25(Wed)10:45:03 No.103994706
Anonymous 01/22/25(Wed)10:45:58 No.103994721
>>103994423
here's .85
here's .85
Anonymous 01/22/25(Wed)10:46:39 No.103994725
>>103994618
i cant tell by the schizoid clusterfuck of metadata comfyui spits out, is this sd3 or flux?
looks fucking crazy man
i cant tell by the schizoid clusterfuck of metadata comfyui spits out, is this sd3 or flux?
looks fucking crazy man
Anonymous 01/22/25(Wed)10:48:02 No.103994741
Anonymous 01/22/25(Wed)10:51:53 No.103994779
How can I load my models from a folder in Metastable? I added the custom folder but the models don't show up.
Anonymous 01/22/25(Wed)11:01:15 No.103994864
Anonymous 01/22/25(Wed)11:01:28 No.103994869
>>103994741
thanks. fuck. i need to upgrade my GPU stat. I wanna contribute to the supply of the uncanny and surreal.
thanks. fuck. i need to upgrade my GPU stat. I wanna contribute to the supply of the uncanny and surreal.
Anonymous 01/22/25(Wed)11:02:22 No.103994880
Anonymous 01/22/25(Wed)11:02:44 No.103994885
Anonymous 01/22/25(Wed)11:03:24 No.103994892
>>103994779
I see you are not very technical. If you were, you would use the word directory instead of folder.
I see you are not very technical. If you were, you would use the word directory instead of folder.
Anonymous 01/22/25(Wed)11:17:20 No.103995078
Anonymous 01/22/25(Wed)11:19:47 No.103995112
>>103994618
trying to gen this using the wf. Had to get the lora, and make a change.
where did you find the fp32? I have an fp8 checkpoint, and I have vanilla dev. I'm genning here with vanilla dev, for no reason.
(not) 162 seconds per iteration: the first iteration takes that long counting load time. It's actually 30 seconds per iteration (sits) on my 6950xt with power budget lowered (I need a new ps).
trying to gen this using the wf. Had to get the lora, and make a change.
where did you find the fp32? I have an fp8 checkpoint, and I have vanilla dev. I'm genning here with vanilla dev, for no reason.
(not) 162 seconds per iteration: the first iteration takes that long counting load time. It's actually 30 seconds per iteration (sits) on my 6950xt with power budget lowered (I need a new ps).
Anonymous 01/22/25(Wed)11:22:27 No.103995153
>>103994725
also why didn't you just grep flux?
@Computer:~/.IGNORE$ grep flux anime\ convention.png
grep: anime convention.png: binary file matches
also why didn't you just grep flux?
@Computer:~/.IGNORE$ grep flux anime\ convention.png
grep: anime convention.png: binary file matches
Anonymous 01/22/25(Wed)11:23:28 No.103995161
YES MY COMPUTER IS NAMED COMPUTER
Anonymous 01/22/25(Wed)11:23:36 No.103995163
>>103995153
He's very retard please understand
He's very retard please understand
Anonymous 01/22/25(Wed)11:26:45 No.103995195
>>103995112
I found it in civitai, had to do some updates on my install procedures because their version there switched to fp32 all the sudden for some fucking reason
https://files.catbox.moe/6nas0q.png
I found it in civitai, had to do some updates on my install procedures because their version there switched to fp32 all the sudden for some fucking reason
https://files.catbox.moe/6nas0q.png
Anonymous 01/22/25(Wed)11:27:05 No.103995199
>>103995112
annnd it's noise, canceled. trying without the lora (maybe I got the wrong version?)
annnd it's noise, canceled. trying without the lora (maybe I got the wrong version?)
Anonymous 01/22/25(Wed)11:27:56 No.103995209
>>103995199
https://civitai.com/api/download/models/983977?type=Model&format=SafeTensor
This is the lora
https://civitai.com/api/download/mo
This is the lora
Anonymous 01/22/25(Wed)11:30:44 No.103995234
Anonymous 01/22/25(Wed)11:32:23 No.103995253
>>103995234
https://civitai.com/api/download/models/691639?type=Model&format=SafeTensor&size=pruned&fp=fp32
tldr pruned
https://civitai.com/api/download/mo
tldr pruned
Anonymous 01/22/25(Wed)11:37:28 No.103995311
I have enabled the brain
Anonymous 01/22/25(Wed)11:39:03 No.103995332
What is it about image generation that draws in absolute retard mongrel psychopaths?
Anonymous 01/22/25(Wed)11:39:57 No.103995339
>>103995311
Model is the big brain file after all.
Model is the big brain file after all.
Anonymous 01/22/25(Wed)11:41:03 No.103995345
>>103995332
You tell me. I bet you're still living in your parents' house.
You tell me. I bet you're still living in your parents' house.
Anonymous 01/22/25(Wed)11:45:52 No.103995404
>>103995253
thanks, downloading. I think my problem is that I need to update comfyui. I think that my version of amdgpu isn't compatible with old comfyui. maybe idk.
btw, you misspelled professional, in the chicken gen. I'm sticking with the misspelling for now.
thanks, downloading. I think my problem is that I need to update comfyui. I think that my version of amdgpu isn't compatible with old comfyui. maybe idk.
btw, you misspelled professional, in the chicken gen. I'm sticking with the misspelling for now.
Anonymous 01/22/25(Wed)11:49:28 No.103995447
>>103995404
it's misspelled in the trigger word instructions that the author made, so I misspelled it too
https://files.catbox.moe/ke33ff.png
it's misspelled in the trigger word instructions that the author made, so I misspelled it too
https://files.catbox.moe/ke33ff.png
Anonymous 01/22/25(Wed)11:49:32 No.103995448
>>103995404
Before updating, make a backup. New CUI versions are getting bloated...
Before updating, make a backup. New CUI versions are getting bloated...
Anonymous 01/22/25(Wed)11:52:41 No.103995493
>>103995448
I think I also need to update pytorch idk for certain, sheesh.
I think I also need to update pytorch idk for certain, sheesh.
Anonymous 01/22/25(Wed)11:55:02 No.103995521
>FETCH ComfyRegistry Data: 15/30
huh
huh
Anonymous 01/22/25(Wed)11:55:31 No.103995529
>>103994418
I am having such a damn difficult time trying to re-create this guy's work.
I am having such a damn difficult time trying to re-create this guy's work.
Anonymous 01/22/25(Wed)11:57:05 No.103995548
>the update seems to have broken preview
Anonymous 01/22/25(Wed)12:01:26 No.103995590
He pulled?
Anonymous 01/22/25(Wed)12:02:05 No.103995598
>He pulled?
>break a feature
>break a feature
Anonymous 01/22/25(Wed)12:07:45 No.103995649
>>103995598
My grandfather, count Orloch!
My grandfather, count Orloch!
Anonymous 01/22/25(Wed)12:16:40 No.103995715
Anonymous 01/22/25(Wed)12:17:40 No.103995728
>>103995598
lmaooo oh well
lmaooo oh well
Anonymous 01/22/25(Wed)12:18:41 No.103995734
Anonymous 01/22/25(Wed)12:22:17 No.103995762
Anonymous 01/22/25(Wed)12:36:57 No.103995915
Anonymous 01/22/25(Wed)12:40:44 No.103995946
soul
Anonymous 01/22/25(Wed)12:42:02 No.103995954
>>103994337
its simple VLM talk to it like youre a chatbot
its simple VLM talk to it like youre a chatbot
Anonymous 01/22/25(Wed)12:43:19 No.103995968
Anonymous 01/22/25(Wed)12:45:24 No.103995993
Anonymous 01/22/25(Wed)12:46:11 No.103996000
Anonymous 01/22/25(Wed)12:46:29 No.103996004
>>103995968
Y the looong face
Y the looong face
Anonymous 01/22/25(Wed)12:48:31 No.103996025
>>103995915
Prove it.
Prove it.
Anonymous 01/22/25(Wed)12:49:26 No.103996038
Anonymous 01/22/25(Wed)12:51:16 No.103996056
>>103996025
That reply is amusing considering denoise is image generation 101
That reply is amusing considering denoise is image generation 101
Anonymous 01/22/25(Wed)12:59:22 No.103996152
Will take a while, but initial results are interesting
Anonymous 01/22/25(Wed)13:06:18 No.103996239
tfw the last anon lora posted here was back in the early Flux days
Anonymous 01/22/25(Wed)13:06:57 No.103996247
Why did Automatic die?
Anonymous 01/22/25(Wed)13:07:55 No.103996258
>>103996056
Empty latent image has nothing to denoise. Denoise as a setting does not matter.
Empty latent image has nothing to denoise. Denoise as a setting does not matter.
Anonymous 01/22/25(Wed)13:08:56 No.103996269
>>103996247
The dev drank himself to death. Not the first such case
The dev drank himself to death. Not the first such case
Anonymous 01/22/25(Wed)13:11:25 No.103996300
How do I teach 3.5 wtf an axolotl is?
Anonymous 01/22/25(Wed)13:12:01 No.103996307
>>103996258
It took you awhile to come up with that meaningless answer. Tell me, if it "does not matter" then why is it a setting in the first place?
But, you're right. An empty latent has nothing to denoise because it does not denoise at all. Why are you here?
It took you awhile to come up with that meaningless answer. Tell me, if it "does not matter" then why is it a setting in the first place?
But, you're right. An empty latent has nothing to denoise because it does not denoise at all. Why are you here?
Anonymous 01/22/25(Wed)13:12:20 No.103996312
>>103996247
Because of Emad, whom is also the reason why SAI died too
Because of Emad, whom is also the reason why SAI died too
Anonymous 01/22/25(Wed)13:13:32 No.103996326
>>103996247
Or big companies not wanting AI to be easily accessible to normies put Ketamine in his beer.
Or big companies not wanting AI to be easily accessible to normies put Ketamine in his beer.
Anonymous 01/22/25(Wed)13:15:07 No.103996351
Anonymous 01/22/25(Wed)13:16:15 No.103996370
>>103996307
Why are you asking?
Why are you asking?
Anonymous 01/22/25(Wed)13:16:27 No.103996372
>>103996351
That's probably fair. I am just trying to make anime images. What do you recommend?
That's probably fair. I am just trying to make anime images. What do you recommend?
Anonymous 01/22/25(Wed)13:17:00 No.103996379
>>103996372
noobai or any illustrious tune
noobai or any illustrious tune
Anonymous 01/22/25(Wed)13:18:17 No.103996393
>>103996307
You type a lot like Ran/Catjak... please stay in your discord.
You type a lot like Ran/Catjak... please stay in your discord.
Anonymous 01/22/25(Wed)13:18:54 No.103996402
Nice double reply
Anonymous 01/22/25(Wed)13:19:17 No.103996408
are there docker images of old comfyui?
Anonymous 01/22/25(Wed)13:20:17 No.103996414
nice double reply
Anonymous 01/22/25(Wed)13:26:39 No.103996502
>>103996402
what do you mean
what do you mean
Anonymous 01/22/25(Wed)13:29:13 No.103996536
>>103996258
Denoise 1.0
Denoise 1.0
Anonymous 01/22/25(Wed)13:29:45 No.103996542
Anonymous 01/22/25(Wed)13:31:34 No.103996566
Anonymous 01/22/25(Wed)13:35:53 No.103996619
>>103996312
SAIs not dead they're busy pumping out 3D models that no one asked for.
SAIs not dead they're busy pumping out 3D models that no one asked for.
Anonymous 01/22/25(Wed)13:43:04 No.103996705
>>103993244
good elf
good elf
Anonymous 01/22/25(Wed)13:43:09 No.103996706
>>103985320
what's the best diffusion model for anime?
what's the best diffusion model for anime?
Anonymous 01/22/25(Wed)13:48:48 No.103996765
>>103996706
Stable Diffusion.
Stable Diffusion.
Anonymous 01/22/25(Wed)13:49:20 No.103996772
>>103996765
This
This
Anonymous 01/22/25(Wed)13:50:51 No.103996796
Anonymous 01/22/25(Wed)13:51:29 No.103996803
Anonymous 01/22/25(Wed)13:52:24 No.103996820
>>103996803
apologies, I cannot read
apologies, I cannot read
Anonymous 01/22/25(Wed)13:55:30 No.103996882
Anonymous 01/22/25(Wed)13:57:20 No.103996910
>>103996706
Well, as for me, personally, I'm going to say CatTower every single time
I feel like I'm in the minority here in this thread, everyone else seems to prefer NoobAI or Illustrious, but as for me, CatTower pumps out banger after banger every single time
Well, as for me, personally, I'm going to say CatTower every single time
I feel like I'm in the minority here in this thread, everyone else seems to prefer NoobAI or Illustrious, but as for me, CatTower pumps out banger after banger every single time
Anonymous 01/22/25(Wed)13:58:28 No.103996929
>>103993192
this, but the loras are still superior most of the time (if you have the storage).
this, but the loras are still superior most of the time (if you have the storage).
Anonymous 01/22/25(Wed)13:59:37 No.103996944
Role? Strategist?
Anonymous 01/22/25(Wed)13:59:40 No.103996945
>>103996929
What lora is that?
What lora is that?
Anonymous 01/22/25(Wed)14:01:04 No.103996972
>>103996536
kino output
kino output
Anonymous 01/22/25(Wed)14:01:37 No.103996982
>>103996945
https://civitai.com/models/547383/eboraart-style-lora-for-ponyxlillustrious
https://danbooru.donmai.us/posts?tags=ebora&z=1
https://civitai.com/models/547383/e
https://danbooru.donmai.us/posts?ta
Anonymous 01/22/25(Wed)14:02:57 No.103997001
>>103996910
thanks, noted
are there any distilled/"turbo"/whatever versions of these? I don't have a GPU
thanks, noted
are there any distilled/"turbo"/whatever versions of these? I don't have a GPU
Anonymous 01/22/25(Wed)14:07:55 No.103997071
Anonymous 01/22/25(Wed)14:11:42 No.103997126
>>103997001
Not that I know of; sorry.
Not that I know of; sorry.
Anonymous 01/22/25(Wed)14:35:47 No.103997432
>>103996408 (me)
https://github.com/hartmark/sd-rocm
this looks nice.
reddiot:
> För example if you have ollama on port 11434 you can just set http://host.docker.internal:11434 in your workflow and that will translate to localhost on docker host.
https://github.com/hartmark/sd-rocm
this looks nice.
reddiot:
> För example if you have ollama on port 11434 you can just set http://host.docker.internal:11434 in your workflow and that will translate to localhost on docker host.
Anonymous 01/22/25(Wed)14:36:56 No.103997450
>>103996536
joke's on you, it's not even a nose
joke's on you, it's not even a nose