/ldg/ - Local Diffusion General
Anonymous 01/19/25(Sun)13:18:24 | 316 comments | 106 images | 🔒 Locked
I'm a drawer, I draw edition
Discussion of Free and Open-Source Diffusion models.
Gone gone once again: >>103944650
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.
Gone gone once again: >>103944650
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 01/19/25(Sun)13:19:01 No.103957388
Blessed thread of frenship
Anonymous 01/19/25(Sun)13:19:06 No.103957389
>Anch- ACK!
Anonymous 01/19/25(Sun)13:19:24 No.103957396
Blessed thread of frenship
Anonymous 01/19/25(Sun)13:20:46 No.103957415
>last thread archived without new bake
Sounds to me like we need to merge with /sdg/.
Sounds to me like we need to merge with /sdg/.
Anonymous 01/19/25(Sun)13:20:47 No.103957416
>my gen didnt make it into the collage
rigged
rigged
Anonymous 01/19/25(Sun)13:21:15 No.103957422
we are so back
Anonymous 01/19/25(Sun)13:22:26 No.103957436
Anonymous 01/19/25(Sun)13:23:05 No.103957440
>>103957415
Not until debo kills himself
Not until debo kills himself
Anonymous 01/19/25(Sun)13:23:49 No.103957451
Double blessed bread :)
Anonymous 01/19/25(Sun)13:24:31 No.103957458
>>103957440
what is it with the man crush on debo?
what is it with the man crush on debo?
Anonymous 01/19/25(Sun)13:25:32 No.103957478
There's only one anons who thinks like >>103957415 btw
Anonymous 01/19/25(Sun)13:25:35 No.103957480
>>103957436
I baked though?
I baked though?
Anonymous 01/19/25(Sun)13:26:57 No.103957501
>>103957480
Where's your bake, I still have the time to delete this one.
Where's your bake, I still have the time to delete this one.
Anonymous 01/19/25(Sun)13:27:27 No.103957504
>>103957501
Uh I baked this thread though?
Uh I baked this thread though?
Anonymous 01/19/25(Sun)13:28:17 No.103957516
For shame.
Anonymous 01/19/25(Sun)13:28:43 No.103957526
Anonymous 01/19/25(Sun)13:31:07 No.103957548
>>103957389
Anchor? We don't really need an anchor thodesu
Anchor? We don't really need an anchor thodesu
Anonymous 01/19/25(Sun)13:34:01 No.103957579
We got trolled.
Anonymous 01/19/25(Sun)13:34:43 No.103957585
>>103957579
Never trust an anon who says they're baking, and never ask who's baking. Just do it.
Never trust an anon who says they're baking, and never ask who's baking. Just do it.
Anonymous 01/19/25(Sun)13:35:21 No.103957594
Anonymous 01/19/25(Sun)13:36:55 No.103957618
Anonymous 01/19/25(Sun)13:37:50 No.103957634
>>103957585
True. If you're baking in good faith, you can always just delete in case it's a doublebake.
True. If you're baking in good faith, you can always just delete in case it's a doublebake.
Anonymous 01/19/25(Sun)13:38:57 No.103957652
Anonymous 01/19/25(Sun)13:41:00 No.103957681
>>103957652
i can post where i like, thanks.
i can post where i like, thanks.
Anonymous 01/19/25(Sun)13:41:28 No.103957693
>>103957681
No you really can't, thanks.
No you really can't, thanks.
Anonymous 01/19/25(Sun)13:43:13 No.103957712
Anonymous 01/19/25(Sun)13:43:42 No.103957722
Who would pretend to bake only to then not bake? Hm... I wonder...
Anonymous 01/19/25(Sun)13:52:12 No.103957843
>>103957618
>>103957712
Thinking about downloading the koff illustrious "A" lora. That one looks the most interesting out of the four.
>>103957712
Thinking about downloading the koff illustrious "A" lora. That one looks the most interesting out of the four.
Anonymous 01/19/25(Sun)13:53:53 No.103957869
>>103957803
this sounds like a question for /lmg/?
this sounds like a question for /lmg/?
Anonymous 01/19/25(Sun)13:54:04 No.103957871
>>103957843
*koff*
*koff*
Anonymous 01/19/25(Sun)13:54:59 No.103957884
I can confirm making a lora for JoyCaption does in fact make it better. Makes me wonder why the creator is such a retard and way too focused on making it generalized for Q&A, as if we need to interrogate the image for how many visible nipples.
Anonymous 01/19/25(Sun)13:55:22 No.103957891
>>103957379
TY for bake <3
TY for bake <3
Anonymous 01/19/25(Sun)13:58:29 No.103957934
Anonymous 01/19/25(Sun)13:58:36 No.103957935
>>103957884
>I can confirm making a lora for JoyCaption does in fact make it better.
How do you make a lora for vision model?
>I can confirm making a lora for JoyCaption does in fact make it better.
How do you make a lora for vision model?
Anonymous 01/19/25(Sun)14:03:23 No.103957991
BANG FOR YOUR BUCK OPTIMIZATION TIP:
This genned in 10 steps. No inpainting, no adetailer. How I did it:
>dmd2_sdxl_4step_lora_fp16
This LORA is NOT just a typical turbo lora. DMD2 speeds SDXL up AND improves quality, consistency, and composition. I learned about this when looking at the comparisons in https://arxiv.org/pdf/2412.02030
>Steps: 10, Sampler: LCM, Schedule type: Align Your Steps, CFG scale: 1.3
DMD2 is made for 4 steps but can go over 10 steps without frying, and quality improves significantly. 8-12 steps is a good range. Align Your Steps is the best scheduler in my XYZ testing. CFG can be 1 (and reduce steps) with high quality output, but also works in low values like 1.1-1.5 if you want a negative prompt. The negative prompt of course increases gen time.
>Smooth detail Booster
A nice detailer LORA for Illustrious models. Not necessary to get benefits from DMD2.
>Emphasis: No norm
Required to stop big prompts from frying on SDXL models
Final tip, these ADetailer settings work well for faces with this setup:
>ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.5, ADetailer dilate erode: 4, ADetailer mask blur: 14, ADetailer denoising strength: 0.25, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer use inpaint width height: True, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer use separate steps: True, ADetailer steps: 7, ADetailer use separate CFG scale: True, ADetailer CFG scale: 1.0
This genned in 10 steps. No inpainting, no adetailer. How I did it:
>dmd2_sdxl_4step_lora_fp16
This LORA is NOT just a typical turbo lora. DMD2 speeds SDXL up AND improves quality, consistency, and composition. I learned about this when looking at the comparisons in https://arxiv.org/pdf/2412.02030
>Steps: 10, Sampler: LCM, Schedule type: Align Your Steps, CFG scale: 1.3
DMD2 is made for 4 steps but can go over 10 steps without frying, and quality improves significantly. 8-12 steps is a good range. Align Your Steps is the best scheduler in my XYZ testing. CFG can be 1 (and reduce steps) with high quality output, but also works in low values like 1.1-1.5 if you want a negative prompt. The negative prompt of course increases gen time.
>Smooth detail Booster
A nice detailer LORA for Illustrious models. Not necessary to get benefits from DMD2.
>Emphasis: No norm
Required to stop big prompts from frying on SDXL models
Final tip, these ADetailer settings work well for faces with this setup:
>ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.5, ADetailer dilate erode: 4, ADetailer mask blur: 14, ADetailer denoising strength: 0.25, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer use inpaint width height: True, ADetailer inpaint width: 768, ADetailer inpaint height: 768, ADetailer use separate steps: True, ADetailer steps: 7, ADetailer use separate CFG scale: True, ADetailer CFG scale: 1.0
Anonymous 01/19/25(Sun)14:04:16 No.103958000
Anonymous 01/19/25(Sun)14:04:45 No.103958006
>>103957935
https://github.com/fpgaminer/joycaption/tree/main/finetuning
His instructions are pretty good. Have to use WSL if you're on Windows, needs 24 GB of VRAM but I'm sure you can do 8-bit training with changing the code. I ended up lowing the rank size by half.
Basically you do:
1. Do a captioning workflow using it. Fix the captions it writes to be the style, tone, length, etc, you like. I found if you gave context with the user prompt you can improve its normal caption a lot.
2. Do a hundred or so images captioned correctly.
3. Use those captions and images to create a JSON file using the described format.
4. Train the Lora.
5. Use the Lora to do more captions, faster.
6. Repeat until the average caption meets your requirements.
https://github.com/fpgaminer/joycap
His instructions are pretty good. Have to use WSL if you're on Windows, needs 24 GB of VRAM but I'm sure you can do 8-bit training with changing the code. I ended up lowing the rank size by half.
Basically you do:
1. Do a captioning workflow using it. Fix the captions it writes to be the style, tone, length, etc, you like. I found if you gave context with the user prompt you can improve its normal caption a lot.
2. Do a hundred or so images captioned correctly.
3. Use those captions and images to create a JSON file using the described format.
4. Train the Lora.
5. Use the Lora to do more captions, faster.
6. Repeat until the average caption meets your requirements.
Anonymous 01/19/25(Sun)14:05:59 No.103958024
Anonymous 01/19/25(Sun)14:06:18 No.103958031
>>103957869
You are absolutely right about that. I had two tabs open and these thread names were too similar.
You are absolutely right about that. I had two tabs open and these thread names were too similar.
Anonymous 01/19/25(Sun)14:09:11 No.103958075
>>103958000
H.R. Giger?
H.R. Giger?
Anonymous 01/19/25(Sun)14:10:52 No.103958093
>>103957991
This gen looks so familiar
This gen looks so familiar
Anonymous 01/19/25(Sun)14:14:33 No.103958132
>>103958075
Yeah! Made a 64 image dataset that needs to be cleaned.
Yeah! Made a 64 image dataset that needs to be cleaned.
Anonymous 01/19/25(Sun)14:15:07 No.103958136
Posting in the extra blessed bread
Anonymous 01/19/25(Sun)14:16:11 No.103958147
Okay smart guys. When do we get O1 like "thinking" diffusion model that iterates on its own work, spots mistakes, corrects them and has some kind of "vision" and executes based on that?
Anonymous 01/19/25(Sun)14:18:49 No.103958159
>>103958147
Are you planning on owning $8000 in GPUs? Also we need to use token-based image AIs if we want this too, it won't work with diffusion.
Are you planning on owning $8000 in GPUs? Also we need to use token-based image AIs if we want this too, it won't work with diffusion.
Anonymous 01/19/25(Sun)14:22:19 No.103958208
>>103958024
The examples on civitai picrel look IMO cooler, but from that grid "B" looks best
>>103958132
Would you mind uploading it for anon? It's really cool regardless
The examples on civitai picrel look IMO cooler, but from that grid "B" looks best
>>103958132
Would you mind uploading it for anon? It's really cool regardless
Anonymous 01/19/25(Sun)14:23:22 No.103958218
>>103957991
I tried dmd2 sdxl lora recently and as expected it didn't perform well with illustrious(-based)+golden noise regardless of the sampler/scheduler combo (the selection of which is already limited by the golden noise). Too comfy to understand what the rest of that means.
I tried dmd2 sdxl lora recently and as expected it didn't perform well with illustrious(-based)+golden noise regardless of the sampler/scheduler combo (the selection of which is already limited by the golden noise). Too comfy to understand what the rest of that means.
Anonymous 01/19/25(Sun)14:33:54 No.103958327
>>103957618
Cute! Wondering, what's the significance of the ghosts in your gens?
Cute! Wondering, what's the significance of the ghosts in your gens?
Anonymous 01/19/25(Sun)14:34:13 No.103958331
>>103957379
Anyone tried experimenting with different settings for gradient_accumulation_steps for HunyuanVid training in diffusion-pipe?
Anyone tried experimenting with different settings for gradient_accumulation_steps for HunyuanVid training in diffusion-pipe?
Anonymous 01/19/25(Sun)14:35:12 No.103958340
>>103958327
The ghosts of the children I kidnapped whilst pushing shopping carts
The ghosts of the children I kidnapped whilst pushing shopping carts
Anonymous 01/19/25(Sun)14:36:03 No.103958349
>>103958331
Gradient accumulation trades speed for VRAM savings. It shouldn't change the final output by much.
Gradient accumulation trades speed for VRAM savings. It shouldn't change the final output by much.
Anonymous 01/19/25(Sun)14:38:53 No.103958375
Anonymous 01/19/25(Sun)14:39:04 No.103958377
>>103958327
i like the aesthetic of ghosts, specifically the round white type, like boo from the mario series. i don't believe in ghosts though.
>>103958340
the cart 'comes in handy'
i like the aesthetic of ghosts, specifically the round white type, like boo from the mario series. i don't believe in ghosts though.
>>103958340
the cart 'comes in handy'
Anonymous 01/19/25(Sun)14:42:29 No.103958409
>>103958349
I see, good info. Thanks
I see, good info. Thanks
Anonymous 01/19/25(Sun)14:48:41 No.103958490
Anonymous 01/19/25(Sun)14:50:16 No.103958507
Anonymous 01/19/25(Sun)14:50:34 No.103958510
How to fix the lifeless eyes problem?
Anonymous 01/19/25(Sun)14:53:57 No.103958546
>>103958510
Visine.
Visine.
Anonymous 01/19/25(Sun)14:55:20 No.103958560
>>103958510
have them look at something, or someone, or closed
have them look at something, or someone, or closed
Anonymous 01/19/25(Sun)15:15:01 No.103958798
>>103958510
Eye positioning is by far the most unbelievable pain in the ass in all of image gen. One of the things for which I most wish I knew how to actually draw as correcting that sort of thing is trivially easy if you have the skills. It is easier when they're looking at something close or down. Newer models can handle directing eyes down/up or to the side and that combined with a little gimp/photoshop can generally pull off what you want.
Eye positioning is by far the most unbelievable pain in the ass in all of image gen. One of the things for which I most wish I knew how to actually draw as correcting that sort of thing is trivially easy if you have the skills. It is easier when they're looking at something close or down. Newer models can handle directing eyes down/up or to the side and that combined with a little gimp/photoshop can generally pull off what you want.
Anonymous 01/19/25(Sun)15:25:28 No.103958903
>>103958507
I gotta clean the dataset, it has 9 buckets
I gotta clean the dataset, it has 9 buckets
Anonymous 01/19/25(Sun)15:26:57 No.103958915
what is pixel smith and how do I use it? using sd.next as an interface right now.
Anonymous 01/19/25(Sun)15:30:27 No.103958950
>>103958006
This is good info, thank you.
This is good info, thank you.
Anonymous 01/19/25(Sun)15:30:45 No.103958956
chicken nuggets
Anonymous 01/19/25(Sun)15:38:28 No.103959041
Anonymous 01/19/25(Sun)15:39:11 No.103959052
Anonymous 01/19/25(Sun)15:39:49 No.103959057
total chinese supremacy
Anonymous 01/19/25(Sun)15:53:26 No.103959206
>>103959057
The algorithm deems you Un mutual.
The algorithm deems you Un mutual.
Anonymous 01/19/25(Sun)15:55:54 No.103959227
Heh nothin personal, kid
Anonymous 01/19/25(Sun)15:56:57 No.103959236
Anon tells me this thread is blessed?
Anonymous 01/19/25(Sun)15:58:07 No.103959250
Do you guys think i2v will be sufficient for specific styles/characters/people or will img2vid + lora be needed?
Anonymous 01/19/25(Sun)16:06:03 No.103959347
>>103959250
I think to truly make great videos we will need image to video + lora + controlnets (video input) + some way of supplying video input (not just images) + some kind of equivalent in video gen to inpainting + some good video editing software. Otherwise its just making memes and small 5 second porn gifs.
I think to truly make great videos we will need image to video + lora + controlnets (video input) + some way of supplying video input (not just images) + some kind of equivalent in video gen to inpainting + some good video editing software. Otherwise its just making memes and small 5 second porn gifs.
Anonymous 01/19/25(Sun)16:10:05 No.103959406
Anonymous 01/19/25(Sun)16:38:35 No.103959743
>>103959250
i2v only makes alternate reality versions of images, Hyvid is smart enough, diverse enough and good enough to probably animate any image you give it.
i2v only makes alternate reality versions of images, Hyvid is smart enough, diverse enough and good enough to probably animate any image you give it.
Anonymous 01/19/25(Sun)16:40:20 No.103959760
>>103959250
Chinese new year just around the corner
Chinese new year just around the corner
Anonymous 01/19/25(Sun)16:41:30 No.103959774
https://www.reddit.com/r/StableDiffusion/comments/1i55bn3/teacache_wavespeed_first_block_cache_comparison/
First block cache is a meme right?
First block cache is a meme right?
Anonymous 01/19/25(Sun)17:21:54 No.103960186
Anonymous 01/19/25(Sun)17:25:20 No.103960211
There are no image/video models to hope for. The only thing we have is fucking i2v, it's so over.
Anonymous 01/19/25(Sun)17:25:35 No.103960213
>>103960186
Her head come off
Her head come off
Anonymous 01/19/25(Sun)17:27:49 No.103960243
>>103960211
>it's so over.
remember anon, we already lived something like that, after the SD3 fiasco we all thought this was the death of imageGen, and then Flux appeared out of nowhere to save the day, and then hunyuan also appeared out of nowhere
>it's so over.
remember anon, we already lived something like that, after the SD3 fiasco we all thought this was the death of imageGen, and then Flux appeared out of nowhere to save the day, and then hunyuan also appeared out of nowhere
Anonymous 01/19/25(Sun)17:28:23 No.103960253
Anonymous 01/19/25(Sun)17:29:01 No.103960258
>>103960213
I was trying to replace her head with a shining light. Like she doesn't have a head, just a radiant glow, though giving her a slight head shape probably would have been better.
I was trying to replace her head with a shining light. Like she doesn't have a head, just a radiant glow, though giving her a slight head shape probably would have been better.
Anonymous 01/19/25(Sun)17:30:30 No.103960269
>>103960213
I make your head come off
I make your head come off
Anonymous 01/19/25(Sun)17:37:54 No.103960358
>>103960258
Cool image
Cool image
Anonymous 01/19/25(Sun)17:50:36 No.103960487
Flux does hands very well mostly, but it seems to have problems with hand holding. The anime models are a bit better at that though you're likely to still be cussing over it.
Anonymous 01/19/25(Sun)17:52:34 No.103960507
>>103960487
>Flux does hands very well mostly, but it seems to have problems with hand holding
that's what happens when you train a model without any humans interacting with each other, I really hope Tencent would reconsider doing a HunyuanDiT2, would blow Flux out of the fucking water
>Flux does hands very well mostly, but it seems to have problems with hand holding
that's what happens when you train a model without any humans interacting with each other, I really hope Tencent would reconsider doing a HunyuanDiT2, would blow Flux out of the fucking water
Anonymous 01/19/25(Sun)17:53:49 No.103960521
>>103957504
FGT
FGT
Anonymous 01/19/25(Sun)18:01:38 No.103960594
https://files.catbox.moe/ywuzv1.mp4
Anonymous 01/19/25(Sun)18:02:45 No.103960608
>>103960594
damn bruh, this was made with hunyuan?
damn bruh, this was made with hunyuan?
Anonymous 01/19/25(Sun)18:04:03 No.103960620
>>103960608
Looks like Kling
Looks like Kling
Anonymous 01/19/25(Sun)18:04:51 No.103960627
>>103960620
based chinks who let people do porn in their APIs, China can't stop winning recently
based chinks who let people do porn in their APIs, China can't stop winning recently
Anonymous 01/19/25(Sun)18:06:56 No.103960654
>>103960627
China's goal is to be AI leader by 2030
China's goal is to be AI leader by 2030
Anonymous 01/19/25(Sun)18:07:54 No.103960669
>>103960654
they will be this year lol, they have no competition, the US and the EU love to shot themselves in the foot, hard to win a race with only one leg
they will be this year lol, they have no competition, the US and the EU love to shot themselves in the foot, hard to win a race with only one leg
Anonymous 01/19/25(Sun)18:19:58 No.103960790
this thread needs alot more widescreen format
Anonymous 01/19/25(Sun)18:27:52 No.103960877
>>103960654
This
>>103960669
They'll probably eek out ahead this year. Retarded Biden basically supercharged their training and inference efficiency by using Nvidia and currency a weapon.
This
>>103960669
They'll probably eek out ahead this year. Retarded Biden basically supercharged their training and inference efficiency by using Nvidia and currency a weapon.
Anonymous 01/19/25(Sun)18:34:47 No.103960942
>>103960877
>Retarded Biden basically supercharged their training and inference efficiency by using Nvidia and currency a weapon.
that's really humiliating for the US, they're the one that put all those GPU restrictions towards China and they'll lose anyway, glad that Trump is back in business, the adults are in the room now
>Retarded Biden basically supercharged their training and inference efficiency by using Nvidia and currency a weapon.
that's really humiliating for the US, they're the one that put all those GPU restrictions towards China and they'll lose anyway, glad that Trump is back in business, the adults are in the room now
Anonymous 01/19/25(Sun)18:42:45 No.103961017
>>103960942
I don't see how trump can fix this.
I don't see how trump can fix this.
Anonymous 01/19/25(Sun)18:44:16 No.103961030
Dawn of he 1st day:
Three days remain until Hunyuan video will not be able to realistically meet its January deadline.
Three days remain until Hunyuan video will not be able to realistically meet its January deadline.
Anonymous 01/19/25(Sun)18:44:32 No.103961033
>>103961017
I think the main reason why the US companies are so terrified to ditch out political correctness and are willing to cuck their models is because they were governed by a leftist cuck government, Trump loves AI and won't harass them like Biden and those leftCuck senate did
I think the main reason why the US companies are so terrified to ditch out political correctness and are willing to cuck their models is because they were governed by a leftist cuck government, Trump loves AI and won't harass them like Biden and those leftCuck senate did
Anonymous 01/19/25(Sun)18:48:58 No.103961072
Anonymous 01/19/25(Sun)19:07:30 No.103961231
Anonymous 01/19/25(Sun)19:16:35 No.103961312
>48GB 4090
https://xcancel.com/main_horse/status/1880554425804320824#m
https://main-horse.github.io/posts/4090-48gb/
Is this the future?
https://xcancel.com/main_horse/stat
https://main-horse.github.io/posts/
Is this the future?
Anonymous 01/19/25(Sun)19:19:05 No.103961335
Anonymous 01/19/25(Sun)19:19:21 No.103961338
Anonymous 01/19/25(Sun)19:20:45 No.103961351
>>103961312
still waiting for china to make fucking 96gb gpu cards for 1000 dollars to kill Nvdia for real
still waiting for china to make fucking 96gb gpu cards for 1000 dollars to kill Nvdia for real
Anonymous 01/19/25(Sun)19:23:23 No.103961379
>>103961312
with these, could it be feasible to chain a couple together with NVlink, or is that no longer a thing? or is it that most stable diffusion models in general cant do multi gpu?
with these, could it be feasible to chain a couple together with NVlink, or is that no longer a thing? or is it that most stable diffusion models in general cant do multi gpu?
Anonymous 01/19/25(Sun)19:24:17 No.103961388
>>103961379
imgfags are too retarded to figure out multigpu
imgfags are too retarded to figure out multigpu
Anonymous 01/19/25(Sun)19:25:08 No.103961397
>>103961379
>most stable diffusion models in general cant do multi gpu?
yeah, there's still no way to do multigpu, which is quite insane when you think about it, I have a 3090 and a 3060 and I'd love to put the extra memory on the 2nd gpu so that I can go for higher resolutions on hunyuan
>most stable diffusion models in general cant do multi gpu?
yeah, there's still no way to do multigpu, which is quite insane when you think about it, I have a 3090 and a 3060 and I'd love to put the extra memory on the 2nd gpu so that I can go for higher resolutions on hunyuan
Anonymous 01/19/25(Sun)19:26:30 No.103961411
>>103961388
we are trying our hardest please understand an be payshent
we are trying our hardest please understand an be payshent
Anonymous 01/19/25(Sun)19:41:11 No.103961519
Anonymous 01/19/25(Sun)19:45:10 No.103961552
>>103961379
>NVlink
That hasn't been relevant to ML stuff for a while, you can easily run LLMs with very little speed loss over multiple GPUs without NVlink, I have no idea why we cant do it with image gen. My second 3090 is just rotting there.
>NVlink
That hasn't been relevant to ML stuff for a while, you can easily run LLMs with very little speed loss over multiple GPUs without NVlink, I have no idea why we cant do it with image gen. My second 3090 is just rotting there.
Anonymous 01/19/25(Sun)19:49:36 No.103961595
Anonymous 01/19/25(Sun)19:52:09 No.103961617
Anonymous 01/19/25(Sun)20:01:51 No.103961706
>>103960627
The funny thing is that the trick many people are doing to get porn in kling is by prompting a Cross in it supposedly, so it confuses it thinking it's a religious thing.
The funny thing is that the trick many people are doing to get porn in kling is by prompting a Cross in it supposedly, so it confuses it thinking it's a religious thing.
Anonymous 01/19/25(Sun)20:08:46 No.103961777
https://huggingface.co/ostris/Flex.1-alpha
Anonymous 01/19/25(Sun)20:11:03 No.103961793
>>103961777
it's flux schnell but smaller, how can it be good?
it's flux schnell but smaller, how can it be good?
Anonymous 01/19/25(Sun)20:21:56 No.103961888
>>103961351
western dog WILL learn
western dog WILL learn
Anonymous 01/19/25(Sun)20:23:31 No.103961910
>>103961793
It can actually be finetuned properly, all the previous fluxes could not, they always bleed or crap out, that's why you only see small finetunes and not big ones that everyone knows about.
So this will be better than anything before even if it's based on schnell
It can actually be finetuned properly, all the previous fluxes could not, they always bleed or crap out, that's why you only see small finetunes and not big ones that everyone knows about.
So this will be better than anything before even if it's based on schnell
Anonymous 01/19/25(Sun)20:25:29 No.103961933
>>103961910
what about flux dedistilled? that one couldn't be finetuned aswell?
what about flux dedistilled? that one couldn't be finetuned aswell?
Anonymous 01/19/25(Sun)20:34:15 No.103962007
is there a way to have hunyuan use for than one gpu to get longer/larger videos?
Anonymous 01/19/25(Sun)20:34:15 No.103962009
>>103961910
no one will care, I'm not trying to be rude but even the most perfect finetune on a 8b flux schnell will never reach the level of flux dev, no one will run that model
no one will care, I'm not trying to be rude but even the most perfect finetune on a 8b flux schnell will never reach the level of flux dev, no one will run that model
Anonymous 01/19/25(Sun)20:35:35 No.103962023
>>103962007
those guys are working on it https://github.com/chengzeyi/Comfy-WaveSpeed?tab=readme-ov-file#comfy-wavespeed
those guys are working on it https://github.com/chengzeyi/Comfy-
Anonymous 01/19/25(Sun)20:44:20 No.103962092
Anonymous 01/19/25(Sun)20:44:34 No.103962094
>>103962023
I see no proof they can follow through with their plans.
I see no proof they can follow through with their plans.
Anonymous 01/19/25(Sun)20:45:58 No.103962106
>>103962094
yeah, but they're the only ones that claim they're working on it, I don't see anyone else working on that even though that's a really important feature
yeah, but they're the only ones that claim they're working on it, I don't see anyone else working on that even though that's a really important feature
Anonymous 01/19/25(Sun)20:48:08 No.103962131
>>103962106
It exists as a node but you have to pay for the service on another repo. It's also integrated into the Hyvid inference code.
It exists as a node but you have to pay for the service on another repo. It's also integrated into the Hyvid inference code.
Anonymous 01/19/25(Sun)20:49:10 No.103962142
>>103962131
>It exists as a node but you have to pay for the service on another repo
so it's useless lol, I'm talking about an open source node everyone can use freely obviously
>It exists as a node but you have to pay for the service on another repo
so it's useless lol, I'm talking about an open source node everyone can use freely obviously
Anonymous 01/19/25(Sun)20:49:16 No.103962144
Anonymous 01/19/25(Sun)20:52:50 No.103962187
Anonymous 01/19/25(Sun)20:54:21 No.103962213
>>103962187
Is this canaan?
Is this canaan?
Anonymous 01/19/25(Sun)20:56:50 No.103962251
Best NoobAI tag?
Anonymous 01/19/25(Sun)20:56:57 No.103962254
>>103962009
wrong, it will reach way above it because we will be able to finetune it to our liking, and not have the same old flux look.
We can get things like noobai etc out of this
wrong, it will reach way above it because we will be able to finetune it to our liking, and not have the same old flux look.
We can get things like noobai etc out of this
Anonymous 01/19/25(Sun)20:57:52 No.103962262
>>103962254
I doubt a 8b model can be finetuned to beat 12b flux dev, maybe I'm wrong but I'm not betting my horses on it
I doubt a 8b model can be finetuned to beat 12b flux dev, maybe I'm wrong but I'm not betting my horses on it
Anonymous 01/19/25(Sun)20:59:14 No.103962273
>>103962262
If it used danbooru/e621 then it would likely be local anime SOTA
If it used danbooru/e621 then it would likely be local anime SOTA
Anonymous 01/19/25(Sun)21:08:40 No.103962361
>>103962213
>>103962187
reminds me of noir
>>103957379
consider adding \vp\NAPT\ to the sister related boards
byeeeee
>>103962187
reminds me of noir
>>103957379
consider adding \vp\NAPT\ to the sister related boards
byeeeee
Anonymous 01/19/25(Sun)21:12:17 No.103962395
Anonymous 01/19/25(Sun)21:18:57 No.103962444
I'm trying to get a character covered in cum, but there's just a bit of cum on her face and chest. I've tried so much yet she just won't get absolutely covered in cum. What to do?
Anonymous 01/19/25(Sun)21:20:57 No.103962459
>>103962444
use a cum lora?
use a cum lora?
Anonymous 01/19/25(Sun)21:21:35 No.103962467
>>103962444
Go to a psychologist
Go to a psychologist
Anonymous 01/19/25(Sun)21:21:42 No.103962469
>>103962395
>yuri
i can promise you due to personal experience (multiple times) that basically nothing good comes from dating\messing\kissing lesbians or bisexual gals kek
>yuri
i can promise you due to personal experience (multiple times) that basically nothing good comes from dating\messing\kissing lesbians or bisexual gals kek
Anonymous 01/19/25(Sun)21:22:55 No.103962482
Anonymous 01/19/25(Sun)21:23:20 No.103962488
>>103962469
I'm not wanting to date lesbians, never got a date in my life, I'm more of a spectator, I just love to watch yuri animes, but like real yuri animes not fucking undertones or yuri bait like Hibike euphonium lol
https://www.youtube.com/watch?v=dLoRO87lu_8
I'm not wanting to date lesbians, never got a date in my life, I'm more of a spectator, I just love to watch yuri animes, but like real yuri animes not fucking undertones or yuri bait like Hibike euphonium lol
https://www.youtube.com/watch?v=dLo
Anonymous 01/19/25(Sun)21:25:02 No.103962503
Anonymous 01/19/25(Sun)21:29:44 No.103962545
Anonymous 01/19/25(Sun)21:32:35 No.103962559
Anonymous 01/19/25(Sun)21:35:36 No.103962588
>>103962559
In public? What a fuckin' whore. She should be locked up in AI jail, there are AI kids nearby!
In public? What a fuckin' whore. She should be locked up in AI jail, there are AI kids nearby!
Anonymous 01/19/25(Sun)21:39:46 No.103962615
>>103962444
whats your prompt?
whats your prompt?
Anonymous 01/19/25(Sun)21:40:50 No.103962621
Anonymous 01/19/25(Sun)21:42:23 No.103962639
>>103962615
you are disgusting
you are disgusting
Anonymous 01/19/25(Sun)21:43:34 No.103962649
>>103962639
can you read? its not cum.
can you read? its not cum.
Anonymous 01/19/25(Sun)21:45:40 No.103962662
>>103962649
still disgusting, I lost my appetite and it's your fault
still disgusting, I lost my appetite and it's your fault
Anonymous 01/19/25(Sun)21:48:49 No.103962686
Anonymous 01/19/25(Sun)21:50:53 No.103962698
is lustify really the best photorealistic still or are there better ones now?
Anonymous 01/19/25(Sun)21:51:06 No.103962700
>>103962588
>>103962588
I know you're joking but thats how most (western) sites are treating photorealistic AI right now, patreon doesn't allow realistic ai content unless its made from a real person with their consent, same with OF and other adult sites.
Reddit is banning users who post AI nsfw content that looks too much like real people, ig (meta) will shadowban you if you don't label realistic AI content.
They fear realistic AI content
>>103962588
I know you're joking but thats how most (western) sites are treating photorealistic AI right now, patreon doesn't allow realistic ai content unless its made from a real person with their consent, same with OF and other adult sites.
Reddit is banning users who post AI nsfw content that looks too much like real people, ig (meta) will shadowban you if you don't label realistic AI content.
They fear realistic AI content
Anonymous 01/19/25(Sun)21:51:23 No.103962703
>>103962662
>still disgusting
you dont even know what it is.
so how it can be disgusting?
I think its all in your head, maybe take a step back and dont be such a schizo
>still disgusting
you dont even know what it is.
so how it can be disgusting?
I think its all in your head, maybe take a step back and dont be such a schizo
Anonymous 01/19/25(Sun)21:52:15 No.103962710
>>103961072
noice
noice
Anonymous 01/19/25(Sun)21:52:31 No.103962712
>>103962703
nigga stfu
nigga stfu
Anonymous 01/19/25(Sun)21:52:42 No.103962716
>>103962621
well duh, if you're a man no lesbian will want to be with you, it's the definition
well duh, if you're a man no lesbian will want to be with you, it's the definition
Anonymous 01/19/25(Sun)21:53:41 No.103962723
>>103962712
I'm not your "nigga", homie.
I'm not your "nigga", homie.
Anonymous 01/19/25(Sun)21:54:23 No.103962727
Anonymous 01/19/25(Sun)21:54:36 No.103962729
happa sexo
Anonymous 01/19/25(Sun)21:55:00 No.103962733
Anonymous 01/19/25(Sun)21:56:34 No.103962748
>>103962727
watch your mouth gay boy
watch your mouth gay boy
Anonymous 01/19/25(Sun)21:56:51 No.103962752
>>103962727
y'all need to chill out before I downvote you
y'all need to chill out before I downvote you
Anonymous 01/19/25(Sun)21:57:17 No.103962760
>>103962748
im gonna fuck you
im gonna fuck you
Anonymous 01/19/25(Sun)21:57:47 No.103962768
Anonymous 01/19/25(Sun)21:57:55 No.103962770
Anonymous 01/19/25(Sun)21:58:29 No.103962776
>>103962770
What's she so happy about? Is it cum?
What's she so happy about? Is it cum?
Anonymous 01/19/25(Sun)22:02:29 No.103962801
>>103962768
>the jannies won't agree with that
the jannies can read so they will agree with it, its not cum. simple.
>the jannies won't agree with that
the jannies can read so they will agree with it, its not cum. simple.
Anonymous 01/19/25(Sun)22:02:32 No.103962802
>>103962776
do you only think of cum faggot?
do you only think of cum faggot?
Anonymous 01/19/25(Sun)22:03:23 No.103962810
>>103962801
enjoy your 3 days ban I guess
enjoy your 3 days ban I guess
Anonymous 01/19/25(Sun)22:05:23 No.103962833
>>103962819
it's slow as fuck
it's slow as fuck
Anonymous 01/19/25(Sun)22:08:21 No.103962863
>>103962810
but I wont get banned since I didnt break any rules.
the image is worksafe. and just because the liquid depicted in the picture reminds (You) of cum does not mean it is supposed to be cum (the image states it is NOT cum).
you should hire a therapist.
but I wont get banned since I didnt break any rules.
the image is worksafe. and just because the liquid depicted in the picture reminds (You) of cum does not mean it is supposed to be cum (the image states it is NOT cum).
you should hire a therapist.
Anonymous 01/19/25(Sun)22:08:27 No.103962864
>>103962819
You're looking for /lmg/ this is /ldg/ but to answer your question you can load an LLM into RAM but it's MUCH slower than VRAM.
You're looking for /lmg/ this is /ldg/ but to answer your question you can load an LLM into RAM but it's MUCH slower than VRAM.
Anonymous 01/19/25(Sun)22:08:49 No.103962870
Anonymous 01/19/25(Sun)22:10:44 No.103962890
>>103962870
Melted icecreme
Melted icecreme
Anonymous 01/19/25(Sun)22:11:02 No.103962893
>>103962870
could be anything, maybe its glue.
could be anything, maybe its glue.
Anonymous 01/19/25(Sun)22:11:25 No.103962896
>>103962870
nerve gas in a liquid state
nerve gas in a liquid state
Anonymous 01/19/25(Sun)22:12:15 No.103962907
Anonymous 01/19/25(Sun)22:13:48 No.103962922
Anonymous 01/19/25(Sun)22:45:18 No.103963185
>>103958510
did you photobash the girls together? increase padding so that both their faces are in frame and inpaint the eyes with a tag like looking at another or something
did you photobash the girls together? increase padding so that both their faces are in frame and inpaint the eyes with a tag like looking at another or something
Anonymous 01/19/25(Sun)22:47:12 No.103963200
coffee fox
Anonymous 01/19/25(Sun)22:51:54 No.103963226
don't make eye contact
Anonymous 01/19/25(Sun)22:58:13 No.103963277
m u p p e t s
Anonymous 01/19/25(Sun)23:01:38 No.103963312
Anonymous 01/19/25(Sun)23:04:13 No.103963331
How would I train an AI to generate pixel art game maps like this? This is a demo from a cloud service, pixellab.ai. I want basically the same thing but local.
My current plan is to train a Flux LoRA on a bunch of Final Fantasy maps, but I'm wondering if anyone has a better idea (I've never trained a LoRA before and don't really know what I'm doing)
My current plan is to train a Flux LoRA on a bunch of Final Fantasy maps, but I'm wondering if anyone has a better idea (I've never trained a LoRA before and don't really know what I'm doing)
Anonymous 01/19/25(Sun)23:11:23 No.103963394
When do we get a local img2vid that can do titty jiggle? That's all I really want
Anonymous 01/19/25(Sun)23:12:16 No.103963405
>>103963394
wait for the chinks to finish the model I guess
https://github.com/Tencent/HunyuanVideo/issues/172#issuecomment-2592676133
>We are working hard on developing an image-to-video model and believe it will be completed soon. However, the open-sourcing process involves some steps that are beyond our control, so we cannot estimate the exact timeline. We believe that it will be available to everyone soon.
wait for the chinks to finish the model I guess
https://github.com/Tencent/HunyuanV
>We are working hard on developing an image-to-video model and believe it will be completed soon. However, the open-sourcing process involves some steps that are beyond our control, so we cannot estimate the exact timeline. We believe that it will be available to everyone soon.
Anonymous 01/19/25(Sun)23:12:48 No.103963413
>>103963331
a lora + guided inpainting with soft inpainting on could do this, although im not sure how perfect the tiles would be
a lora + guided inpainting with soft inpainting on could do this, although im not sure how perfect the tiles would be
Anonymous 01/19/25(Sun)23:18:19 No.103963452
>>103963405
i2v bros it's over
i2v bros it's over
Anonymous 01/19/25(Sun)23:19:39 No.103963464
>>103962686
Catbox? Love this
Catbox? Love this
Anonymous 01/19/25(Sun)23:33:08 No.103963541
>>103962251
pc-98 \(style\)
pc-98 \(style\)
Anonymous 01/19/25(Sun)23:33:25 No.103963543
>>103963405
>model and believe it will be completed soon. However, the open-sourcing process involves some steps that are beyond our control, so we cannot estimate the exact timeline. We believe that it will be available to everyone soon.
A few questions to gleam from that response.
1. Is the training finished or is it not?
2. What is this open sourcing process they are referring to and why might it be an obstacle?
3. Who is this man and how familiar is he really with the training and release process?
>model and believe it will be completed soon. However, the open-sourcing process involves some steps that are beyond our control, so we cannot estimate the exact timeline. We believe that it will be available to everyone soon.
A few questions to gleam from that response.
1. Is the training finished or is it not?
2. What is this open sourcing process they are referring to and why might it be an obstacle?
3. Who is this man and how familiar is he really with the training and release process?
Anonymous 01/19/25(Sun)23:35:34 No.103963558
>>103963543
>Is the training finished or is it not?
no, if they said "it will be completed soon" it means it's not fully complete/trained
>2. What is this open sourcing process they are referring to and why might it be an obstacle?
the US, the EU, the CPP? a shit ton of governments are putting restrictions on models, that's why Sora isn't allowed to the EU for example
>3. Who is this man and how familiar is he really with the training and release process?
he's an Algorithm Engineer in Tencent
>Is the training finished or is it not?
no, if they said "it will be completed soon" it means it's not fully complete/trained
>2. What is this open sourcing process they are referring to and why might it be an obstacle?
the US, the EU, the CPP? a shit ton of governments are putting restrictions on models, that's why Sora isn't allowed to the EU for example
>3. Who is this man and how familiar is he really with the training and release process?
he's an Algorithm Engineer in Tencent
Anonymous 01/19/25(Sun)23:37:16 No.103963568
Anonymous 01/19/25(Sun)23:37:16 No.103963569
Anonymous 01/19/25(Sun)23:37:18 No.103963570
>>103963558
>if they said "it will be completed soon" it means it's not fully complete/trained
Does it really though? How familiar is this guy with English? Was it a machine translation is the model the only consideration when it comes to calling it complete?
>if they said "it will be completed soon" it means it's not fully complete/trained
Does it really though? How familiar is this guy with English? Was it a machine translation is the model the only consideration when it comes to calling it complete?
Anonymous 01/19/25(Sun)23:38:39 No.103963577
>>103963413
>a lora + guided inpainting with soft inpainting on could do this
I looked up "soft inpainting" and found some info, but what do you mean by "guided inpainting"?
>although im not sure how perfect the tiles would be
That part I'm not too worried about - I don't mind cleaning them up by hand at the end
>a lora + guided inpainting with soft inpainting on could do this
I looked up "soft inpainting" and found some info, but what do you mean by "guided inpainting"?
>although im not sure how perfect the tiles would be
That part I'm not too worried about - I don't mind cleaning them up by hand at the end
Anonymous 01/19/25(Sun)23:39:10 No.103963579
>>103963570
he said they are "developping" the model and that it "will be completed soon", a finished model wouldn't need more developping right?
he said they are "developping" the model and that it "will be completed soon", a finished model wouldn't need more developping right?
Anonymous 01/19/25(Sun)23:39:34 No.103963582
>>103963558
yeah imagefx for example, SORRY NOT AVAILABLE in my shittobeans country. fuck that shit
yeah imagefx for example, SORRY NOT AVAILABLE in my shittobeans country. fuck that shit
Anonymous 01/19/25(Sun)23:40:19 No.103963586
Anonymous 01/19/25(Sun)23:41:45 No.103963593
>>103963579
idk, I don't take ESLs statements at face value without further clarification. I work with ESLs all day and it's a built in reflex of mine to confirm exactly what they mean.
idk, I don't take ESLs statements at face value without further clarification. I work with ESLs all day and it's a built in reflex of mine to confirm exactly what they mean.
Anonymous 01/19/25(Sun)23:42:31 No.103963599
>>103963586
why is your (cool) image 1.6mb? compare that to the one I posted above, same size. 376kb
why is your (cool) image 1.6mb? compare that to the one I posted above, same size. 376kb
Anonymous 01/19/25(Sun)23:44:27 No.103963614
>>103963593
nothing's preventing you to ask further questions on the github
nothing's preventing you to ask further questions on the github
Anonymous 01/19/25(Sun)23:45:10 No.103963616
Anonymous 01/19/25(Sun)23:46:34 No.103963625
The quicker you give up on i2v the eaiser this will be
Anonymous 01/19/25(Sun)23:46:35 No.103963626
>>103963599
jpg compression quality. high compression = lower size but it's also lossy. Yours is highly compressed. Mine, not so much.
jpg compression quality. high compression = lower size but it's also lossy. Yours is highly compressed. Mine, not so much.
Anonymous 01/19/25(Sun)23:46:48 No.103963632
>>103963577
>but what do you mean by "guided inpainting"?
guided inpainting means you do a simple doodle of what you want and inpaint that, like what's being done in your gif
>soft inpainting
soft inpainting is important if you want things to blend together nicely, notice how in your gif it does small changes to the already inpainted bits to help blend in the new content seamlessly? that's what soft inpainting + a higher mask blur (i'd say around 6-10+) does. the more you increase your mask blur value the more small changes it does outside of your main mask. it takes a bit of getting used to but learning how to properly use soft inpainting is essential
>but what do you mean by "guided inpainting"?
guided inpainting means you do a simple doodle of what you want and inpaint that, like what's being done in your gif
>soft inpainting
soft inpainting is important if you want things to blend together nicely, notice how in your gif it does small changes to the already inpainted bits to help blend in the new content seamlessly? that's what soft inpainting + a higher mask blur (i'd say around 6-10+) does. the more you increase your mask blur value the more small changes it does outside of your main mask. it takes a bit of getting used to but learning how to properly use soft inpainting is essential
Anonymous 01/19/25(Sun)23:50:48 No.103963658
Anonymous 01/19/25(Sun)23:51:54 No.103963667
>>103963614
I don't know how that would help me other than reconfirming the fact its not ready yet.
I don't know how that would help me other than reconfirming the fact its not ready yet.
Anonymous 01/19/25(Sun)23:52:09 No.103963669
Anonymous 01/19/25(Sun)23:52:50 No.103963679
>>103963667
I thought you wanted to know about it, seems like it's not that important, and desu I agree with that, that won't change the date we'll get that model
I thought you wanted to know about it, seems like it's not that important, and desu I agree with that, that won't change the date we'll get that model
Anonymous 01/19/25(Sun)23:54:11 No.103963686
Anonymous 01/19/25(Sun)23:56:01 No.103963703
Anonymous 01/19/25(Sun)23:56:02 No.103963704
>>103963679
I guess I naively thought someone here might have a better insight. I realize now that was stupid.
I guess I naively thought someone here might have a better insight. I realize now that was stupid.
Anonymous 01/19/25(Sun)23:58:34 No.103963715
>>103963704
they have a discord, and I went there to find more informations, but all I saw was some troon who was offended that Hunyuan couldn't do troons proprely lol
they have a discord, and I went there to find more informations, but all I saw was some troon who was offended that Hunyuan couldn't do troons proprely lol
Anonymous 01/19/25(Sun)23:59:35 No.103963723
>discord
Sounds about right for discord.
Sounds about right for discord.
Anonymous 01/20/25(Mon)00:01:42 No.103963731
>>103963715
>troon
>offended
find a better duo, even KobexShaq can't beat that
https://www.youtube.com/watch?v=8Jc6jiyqYPQ
>troon
>offended
find a better duo, even KobexShaq can't beat that
https://www.youtube.com/watch?v=8Jc
Anonymous 01/20/25(Mon)00:04:16 No.103963755
Anonymous 01/20/25(Mon)00:06:57 No.103963778
Anonymous 01/20/25(Mon)00:08:03 No.103963788
Anonymous 01/20/25(Mon)00:09:55 No.103963798
>>103963788
He's off to get the img2vid model. Overnight drive from France to China.
He's off to get the img2vid model. Overnight drive from France to China.
Anonymous 01/20/25(Mon)00:11:03 No.103963805
Anonymous 01/20/25(Mon)00:11:34 No.103963810
>>103963798
i hope he doesnt crash on the way ;_;
i hope he doesnt crash on the way ;_;
Anonymous 01/20/25(Mon)00:17:50 No.103963847
Anyone know how I prevent this weird creasing under the boobs with tops?
>please excuse the pigface
>please excuse the pigface
Anonymous 01/20/25(Mon)00:19:01 No.103963860
>>103963847
uhhhhhhhhhhhhhhhhhhhhh inpaint
uhhhhhhhhhhhhhhhhhhhhh inpaint
Anonymous 01/20/25(Mon)00:19:23 No.103963863
>>103963847
use a compatible resolution, use a better model, inpaint, yada yada
use a compatible resolution, use a better model, inpaint, yada yada
Anonymous 01/20/25(Mon)00:21:17 No.103963879
>>103963847
the gen is perfectly fine, face not too bad either. "no worries"
the gen is perfectly fine, face not too bad either. "no worries"
Anonymous 01/20/25(Mon)00:22:50 No.103963892
What happened here?
Anonymous 01/20/25(Mon)00:29:53 No.103963946
Anonymous 01/20/25(Mon)00:38:22 No.103964001
Anonymous 01/20/25(Mon)00:44:47 No.103964047
Anonymous 01/20/25(Mon)00:48:24 No.103964070
Anonymous 01/20/25(Mon)00:56:06 No.103964120
Anonymous 01/20/25(Mon)01:12:14 No.103964227
The tencent twitter fags are alive again
https://xcancel.com/TXhunyuan/status/1881221843920748742#m
>It won't be long.
https://xcancel.com/TXhunyuan/status/1881221568510267794#m
>话虽如此,但是图生视频是不是1月份会开源呢?
>That said, isn't Tencent Video going to be open source in January?
>年后回来,大哥
>Come back in a year, brother.
I used deepL so maybe that's wrong, is there some actual chinks that can give a good translation of this?
https://xcancel.com/TXhunyuan/statu
>It won't be long.
https://xcancel.com/TXhunyuan/statu
>话虽如此,但是图生视频是不是1月份会开源呢?
>That said, isn't Tencent Video going to be open source in January?
>年后回来,大哥
>Come back in a year, brother.
I used deepL so maybe that's wrong, is there some actual chinks that can give a good translation of this?
Anonymous 01/20/25(Mon)01:15:14 No.103964251
Anonymous 01/20/25(Mon)01:16:02 No.103964260
>>103964227
i can feel it, the hunyuan guys are about to do an unfathomable troll
i can feel it, the hunyuan guys are about to do an unfathomable troll
Anonymous 01/20/25(Mon)01:16:35 No.103964265
>>103964251
the "come back in a year, brother" is concerning though
the "come back in a year, brother" is concerning though
Anonymous 01/20/25(Mon)01:19:13 No.103964285
Anonymous 01/20/25(Mon)01:19:48 No.103964291
lustify or the realism by stable yogi, which is better?
Anonymous 01/20/25(Mon)01:20:14 No.103964297
>>103964265
>"come back in a year, brother" is concerning though
Only until you realize Chinese new year is next week.
>"come back in a year, brother" is concerning though
Only until you realize Chinese new year is next week.
Anonymous 01/20/25(Mon)01:22:24 No.103964312
Anonymous 01/20/25(Mon)01:24:42 No.103964329
>>103964312
That has always been my interpretation of the way things are going. They can't release it any later than Wednesday this week. Once the Chinese new year celebration starts most people are not working for the duration. It may as well be Christmas vacation.
That has always been my interpretation of the way things are going. They can't release it any later than Wednesday this week. Once the Chinese new year celebration starts most people are not working for the duration. It may as well be Christmas vacation.
Anonymous 01/20/25(Mon)01:34:48 No.103964410
Anonymous 01/20/25(Mon)01:36:12 No.103964422
Anonymous 01/20/25(Mon)01:43:25 No.103964482
Anonymous 01/20/25(Mon)01:45:44 No.103964504
>>103964422
going to have to update it soon, the man got a hair transplant with his patreonbux
going to have to update it soon, the man got a hair transplant with his patreonbux
Anonymous 01/20/25(Mon)01:53:49 No.103964574
>>103964504
Can I get a source on this?
Can I get a source on this?
Anonymous 01/20/25(Mon)02:01:25 No.103964641
Anonymous 01/20/25(Mon)02:02:29 No.103964654
Anonymous 01/20/25(Mon)02:07:46 No.103964691
>>103964291
lustify all the way, thrillustrious also pretty good
lustify all the way, thrillustrious also pretty good
Anonymous 01/20/25(Mon)02:11:57 No.103964717
Anonymous 01/20/25(Mon)02:21:56 No.103964783
>>103964227
It means new year not in a year in reference to the Chinese New Year wood snake edition. Tencent is closed down for this holiday like many other businesses.
It means new year not in a year in reference to the Chinese New Year wood snake edition. Tencent is closed down for this holiday like many other businesses.
Anonymous 01/20/25(Mon)02:28:46 No.103964820
Anonymous 01/20/25(Mon)02:29:47 No.103964827
in honor of china
Anonymous 01/20/25(Mon)02:49:04 No.103964928
Anonymous 01/20/25(Mon)03:03:55 No.103965031
>>103964820
I cant believe you weren't trolling/
I cant believe you weren't trolling/
Anonymous 01/20/25(Mon)03:13:08 No.103965106
>>103964717
Super clean, is there a specific lora for that?
Super clean, is there a specific lora for that?
Anonymous 01/20/25(Mon)03:20:25 No.103965146
>>103965106
No, not really. I'm using the SPO SDXL lora to clean things up a bit, along with PAG and SEG.
I think part of it is that the model itself (WAI-NSFW-illustrious-SDXL) seems to be able to make more clean/less cluttered images than other models.
No, not really. I'm using the SPO SDXL lora to clean things up a bit, along with PAG and SEG.
I think part of it is that the model itself (WAI-NSFW-illustrious-SDXL) seems to be able to make more clean/less cluttered images than other models.
Anonymous 01/20/25(Mon)03:34:14 No.103965220
Every time I prompt "walking" my 1girl walks directly towards the camera
I want to generate a character walking at an angle. I know it's called a "three quarter view" but stable diffusion doesn't seem to understand that term.
What prompt should I use to get a character walking at an angle?
>Put "walking toward viewer" in negative prompt
Already tried that, doesn't want to work.
I want to generate a character walking at an angle. I know it's called a "three quarter view" but stable diffusion doesn't seem to understand that term.
What prompt should I use to get a character walking at an angle?
>Put "walking toward viewer" in negative prompt
Already tried that, doesn't want to work.
Anonymous 01/20/25(Mon)03:43:10 No.103965274
>>103965220
try "from side" or put "facing viewer" into the negs
try "from side" or put "facing viewer" into the negs
Anonymous 01/20/25(Mon)04:00:22 No.103965386
>>103965274
I appreciate the suggestion but "from side" makes her rotate 90 degrees, even if "facing viewer" isn't in the negs
"from angle" sometimes works, but almost always rotates her into a "dutch angle" shot, even if "dutch angle" is in the negs
All of these anime models were trained off of booru images and tags, but very few images were ever tagged with any kind of term that denotes they're at an angle, so this is tough...
I appreciate the suggestion but "from side" makes her rotate 90 degrees, even if "facing viewer" isn't in the negs
"from angle" sometimes works, but almost always rotates her into a "dutch angle" shot, even if "dutch angle" is in the negs
All of these anime models were trained off of booru images and tags, but very few images were ever tagged with any kind of term that denotes they're at an angle, so this is tough...
Anonymous 01/20/25(Mon)04:06:34 No.103965432
>I can't help it, brother. It's too late.
It's never been more over.
It's never been more over.
Anonymous 01/20/25(Mon)04:25:04 No.103965536
:]
Anonymous 01/20/25(Mon)04:32:41 No.103965561
>>103965536
Most effort poster in this thread by far.
Most effort poster in this thread by far.
Anonymous 01/20/25(Mon)04:33:54 No.103965567
Anonymous 01/20/25(Mon)04:57:06 No.103965693
Anonymous 01/20/25(Mon)05:00:09 No.103965704
>>103965567
'straight-on' in negative. 'three quarter view' and 'three quarter profile' in positive. 'facing viewer' relates to head, not body, position. Sift through the Danbooru tags at the source to find the ones you need.
'straight-on' in negative. 'three quarter view' and 'three quarter profile' in positive. 'facing viewer' relates to head, not body, position. Sift through the Danbooru tags at the source to find the ones you need.
Anonymous 01/20/25(Mon)05:00:14 No.103965706
>>103965567
Not him but have you tried 50/50 weight of facing forward/facing away?
Not him but have you tried 50/50 weight of facing forward/facing away?
Anonymous 01/20/25(Mon)05:05:35 No.103965739
do you guys know of any applications that I could use to outpaint for free on apple silicon? Doesn't need to do much other than add a bunch of grass or sky
Anonymous 01/20/25(Mon)05:22:40 No.103965860
>>103965432
https://xcancel.com/TXhunyuan/status/188126>话虽如此,但是图生视频是不是1月份会开源呢?
>That said, isn't Tencent Video going to be open source in January?
>年后回来,大哥
>Come back in a year, brother
>别年后啊啊啊
Don't "year after" ah ah ah
>没办法啊,大哥,年前来不及
>I can't help it, brother. It's too late.
So basically the release is gonna be after the 29th of January, that's when the new year comes in China
https://xcancel.com/TXhunyuan/statu
>That said, isn't Tencent Video going to be open source in January?
>年后回来,大哥
>Come back in a year, brother
>别年后啊啊啊
Don't "year after" ah ah ah
>没办法啊,大哥,年前来不及
>I can't help it, brother. It's too late.
So basically the release is gonna be after the 29th of January, that's when the new year comes in China
Anonymous 01/20/25(Mon)05:24:20 No.103965874
>>103965860
>Come back in a year
NO I2V FOR YOU, COME BACK ONE YEAR
https://www.youtube.com/watch?v=zOpfsGrNvnk
>Come back in a year
NO I2V FOR YOU, COME BACK ONE YEAR
https://www.youtube.com/watch?v=zOp
Anonymous 01/20/25(Mon)05:28:09 No.103965908
Anonymous 01/20/25(Mon)05:33:16 No.103965949
>>103965860
bruh...
bruh...
Anonymous 01/20/25(Mon)05:35:49 No.103965971
>>103965860
>>103965949
I hope they made the "open release process" with the regulation shit with countries during the break period, so that the waiting will be during the time they won't work
>>103965949
I hope they made the "open release process" with the regulation shit with countries during the break period, so that the waiting will be during the time they won't work
Anonymous 01/20/25(Mon)05:38:35 No.103966002
Anonymous 01/20/25(Mon)05:43:07 No.103966036
>>103965031
>A turkish guy getting a hair transplant
>Turkey is considered the best country when it comes to getting good hair transplant for cheap
lul
>A turkish guy getting a hair transplant
>Turkey is considered the best country when it comes to getting good hair transplant for cheap
lul
Anonymous 01/20/25(Mon)05:59:42 No.103966141
This >>103963599 >>103963626 reminds me:
I decided to stash comfyui/auto generated images as JPEG XL because it's clearly smaller even with lossless compression. Saves a good bit of space, 30-50% ish depending on the image, it depends.
If someone wants to do something comparable with xl converter or any other tool that can use exiftool, add "-m -usercomment<parameters" to preserve the metadata from the png, it otherwise strips those non-standard png text "parameters".
I decided to stash comfyui/auto generated images as JPEG XL because it's clearly smaller even with lossless compression. Saves a good bit of space, 30-50% ish depending on the image, it depends.
If someone wants to do something comparable with xl converter or any other tool that can use exiftool, add "-m -usercomment<parameters" to preserve the metadata from the png, it otherwise strips those non-standard png text "parameters".
Anonymous 01/20/25(Mon)06:02:59 No.103966161
>>103966141
Aw man, I was going to switch to .jxl for my hoarding but then they removed(!) the native support from the browsers.
Aw man, I was going to switch to .jxl for my hoarding but then they removed(!) the native support from the browsers.
Anonymous 01/20/25(Mon)06:07:09 No.103966189
>>103966161
Actually I think the Apple ecosystem added support including the browser, and more importantly most open sauce libs and tools including my image viewers/editors added support. Works fine for me.
I can even shitpost jxl on some other platforms.
Actually I think the Apple ecosystem added support including the browser, and more importantly most open sauce libs and tools including my image viewers/editors added support. Works fine for me.
I can even shitpost jxl on some other platforms.
Anonymous 01/20/25(Mon)06:08:04 No.103966196
>>103957415
we'd let this place burn before ever associating with those dysgenic freaks
we'd let this place burn before ever associating with those dysgenic freaks
Anonymous 01/20/25(Mon)06:13:03 No.103966220
>>103966189
Well yeah, I know, but for me the fact that they actually removed it from Chrome and it didn't even make it into FF, regardless of the extensions, made me think it's not going to take off after all. Which is too bad, I wanted it to succeed and bury webp.
Well yeah, I know, but for me the fact that they actually removed it from Chrome and it didn't even make it into FF, regardless of the extensions, made me think it's not going to take off after all. Which is too bad, I wanted it to succeed and bury webp.
Anonymous 01/20/25(Mon)06:16:51 No.103966240
Anonymous 01/20/25(Mon)06:22:45 No.103966275
>>103966220
I don't know when a threshold for "take off" is met but it works fine for me to store/view the same images with smaller files. The image editors / viewers / other tools almost all picked it up.
There are quite many browsers that support it too though apart from safari I guess you have to pick a fork of ff/chrome rather than use the main version you're probably using.
I'm not sure if it will "bury" other formats, it justwerks for me though. Don't need to pick which files I want in JPEG XL and which in PNG or JPG, it can all be JPEG XL no problem.
I don't know when a threshold for "take off" is met but it works fine for me to store/view the same images with smaller files. The image editors / viewers / other tools almost all picked it up.
There are quite many browsers that support it too though apart from safari I guess you have to pick a fork of ff/chrome rather than use the main version you're probably using.
I'm not sure if it will "bury" other formats, it justwerks for me though. Don't need to pick which files I want in JPEG XL and which in PNG or JPG, it can all be JPEG XL no problem.
Anonymous 01/20/25(Mon)06:32:24 No.103966319
>>103966275
Linking a forma explanation post in case a random anon takes interest https://liliththecatpire.vivaldi.net/2024/12/26/jpeg-xl-is-here-stop-ignoring-it/
But if you can guess my browser from it, alas my only choice for the web would be an extension.
And I know I said my interest was for the purpose of converting the local files, but being stuck with something I can't send or post expecting it to just work is meh; image editors/viewers support is only "almost".
Stupid excuses but what can I do.
Linking a forma explanation post in case a random anon takes interest https://liliththecatpire.vivaldi.ne
But if you can guess my browser from it, alas my only choice for the web would be an extension.
And I know I said my interest was for the purpose of converting the local files, but being stuck with something I can't send or post expecting it to just work is meh; image editors/viewers support is only "almost".
Stupid excuses but what can I do.
Anonymous 01/20/25(Mon)06:34:22 No.103966332
>>103965739
anyone?
anyone?
Anonymous 01/20/25(Mon)06:59:37 No.103966500
>>103966319
Sure, you decide if/when it works for you. It does work for me.
> if you can guess my browser from it
I can't. But if it's any of the common browsers they obviously announce it outright to every site you visit. I'm pretty sure I want to stick with the forks either way because chrome and firefox are both fucking around too much. So I get JXL support no problem.
>>103965739 >>103966332
I don't know Apple. Can you use Krita? YT comes up with this: https://www.youtube.com/watch?v=QMEIiIe_hhM
Sure, you decide if/when it works for you. It does work for me.
> if you can guess my browser from it
I can't. But if it's any of the common browsers they obviously announce it outright to every site you visit. I'm pretty sure I want to stick with the forks either way because chrome and firefox are both fucking around too much. So I get JXL support no problem.
>>103965739 >>103966332
I don't know Apple. Can you use Krita? YT comes up with this: https://www.youtube.com/watch?v=QME
Anonymous 01/20/25(Mon)07:11:32 No.103966579
>>103966500
:^)
:^)
Anonymous 01/20/25(Mon)07:24:34 No.103966688
Anonymous 01/20/25(Mon)07:41:17 No.103966869
Anonymous 01/20/25(Mon)07:59:46 No.103967023
I think it's finaly tiome to move on from A1111 toforge. Is reForge worth it?
Anonymous 01/20/25(Mon)08:01:14 No.103967039
Anonymous 01/20/25(Mon)08:01:37 No.103967045
>>103967023
Perhaps just use https://github.com/LykosAI/StabilityMatrix to test what works for you, it's relatively convenient since you can have it install stuff and share the model checkpoints.
Perhaps just use https://github.com/LykosAI/Stabilit
Anonymous 01/20/25(Mon)08:03:51 No.103967068
>>103967023
very much so anon, go for it.
very much so anon, go for it.
Anonymous 01/20/25(Mon)08:10:22 No.103967130
>>103966688
really like the texure wizardman, you were using pixelwave flux, right? mind sharing your upscaling details? I'm curious.
really like the texure wizardman, you were using pixelwave flux, right? mind sharing your upscaling details? I'm curious.
Anonymous 01/20/25(Mon)08:15:13 No.103967178
Is there a reason to use forge when swarm exists?
Anonymous 01/20/25(Mon)08:21:44 No.103967223
>>103967068
>>103967039
>>103967039
ok I did the git merge to change from a1111 to reforge and nothing changed it's still the old automatic1111
>>103967039
>>103967039
ok I did the git merge to change from a1111 to reforge and nothing changed it's still the old automatic1111
Anonymous 01/20/25(Mon)08:22:37 No.103967231
https://www.youtube.com/watch?v=_Y3_UsxDsLA
When will local reach that level?
When will local reach that level?
Anonymous 01/20/25(Mon)08:44:47 No.103967411
Anonymous 01/20/25(Mon)08:46:27 No.103967436
Anonymous 01/20/25(Mon)08:55:23 No.103967519
>>103967231
no clue, some of this probably is feasible soon™
no clue, some of this probably is feasible soon™
Anonymous 01/20/25(Mon)09:21:52 No.103967746
>>103967130
I had to go to work but I will answer your question this evening when I get back in
I had to go to work but I will answer your question this evening when I get back in
Anonymous 01/20/25(Mon)09:37:08 No.103967870
Anonymous 01/20/25(Mon)09:59:45 No.103968086
>>103968055
>The US said no GPU for China and China answered with no AI money for the US.
on the llm space the chinks released an open model that's as good as the best API models for 50x cheaper, is this why they're releasing great models openly? because they hate the US because of the gpu restriction and as an act of revenge they want to destroy any kind of AI API business or something? if yes that's so fucking based
https://www.chinatalk.media/p/deepseek-ceo-interview-with-chinas
>Liang Wenfeng: In the face of disruptive technologies, moats created by closed source are temporary. Even OpenAI’s closed source approach can’t prevent others from catching up. So we anchor our value in our team — our colleagues grow through this process, accumulate know-how, and form an organization and culture capable of innovation. That’s our moat.
>Open source, publishing papers, in fact, do not cost us anything. For technical talent, having others follow your innovation gives a great sense of accomplishment. In fact, open source is more of a cultural behavior than a commercial one, and contributing to it earns us respect. There is also a cultural attraction for a company to do this.
>The US said no GPU for China and China answered with no AI money for the US.
on the llm space the chinks released an open model that's as good as the best API models for 50x cheaper, is this why they're releasing great models openly? because they hate the US because of the gpu restriction and as an act of revenge they want to destroy any kind of AI API business or something? if yes that's so fucking based
https://www.chinatalk.media/p/deeps
>Liang Wenfeng: In the face of disruptive technologies, moats created by closed source are temporary. Even OpenAI’s closed source approach can’t prevent others from catching up. So we anchor our value in our team — our colleagues grow through this process, accumulate know-how, and form an organization and culture capable of innovation. That’s our moat.
>Open source, publishing papers, in fact, do not cost us anything. For technical talent, having others follow your innovation gives a great sense of accomplishment. In fact, open source is more of a cultural behavior than a commercial one, and contributing to it earns us respect. There is also a cultural attraction for a company to do this.
Anonymous 01/20/25(Mon)10:34:21 No.103968463
>>103967411
froggieman
froggieman
Anonymous 01/20/25(Mon)10:36:17 No.103968484
I made this ages ago in 1.5 with QR monster, is there anything that has been done in SDXL based models that works as well? The ported qrmonster thing on civitai seems bad
Anonymous 01/20/25(Mon)10:51:54 No.103968628
>>103968484
are you refering to "qr code monster", the controlnet? if so, hm. there is a thingie called "x-adapter", you could try that. as in, use an sd1.5 controlnet with sdxl. or you could use deepshrink and gen at a higher base resolution with sd1.5
are you refering to "qr code monster", the controlnet? if so, hm. there is a thingie called "x-adapter", you could try that. as in, use an sd1.5 controlnet with sdxl. or you could use deepshrink and gen at a higher base resolution with sd1.5
Anonymous 01/20/25(Mon)11:06:58 No.103968762
Anonymous 01/20/25(Mon)11:12:41 No.103968808
Anonymous 01/20/25(Mon)11:38:43 No.103969069
Anonymous 01/20/25(Mon)11:46:18 No.103969137
>>103968086
yeah it's based that Chinese foreigners destroyed our jobs, houses and property so you can gen lewds in your depressed dystopia in a destroyed culture prominently eroded by a Chinese propaganda app
yeah it's based that Chinese foreigners destroyed our jobs, houses and property so you can gen lewds in your depressed dystopia in a destroyed culture prominently eroded by a Chinese propaganda app
Anonymous 01/20/25(Mon)11:48:29 No.103969155
>>103969137
>Chinese foreigners destroyed our jobs, houses and property
*indians
>in a destroyed culture
you can thank the leftist wokes for that
>Chinese foreigners destroyed our jobs, houses and property
*indians
>in a destroyed culture
you can thank the leftist wokes for that
Anonymous 01/20/25(Mon)11:50:28 No.103969168
>>103969155
If TikTok never existed our society would be infinitely better off. If Chinese couldn't buy our houses and business, you would have a job and a house. But at least you have a shitty AI model.
If TikTok never existed our society would be infinitely better off. If Chinese couldn't buy our houses and business, you would have a job and a house. But at least you have a shitty AI model.
Anonymous 01/20/25(Mon)11:51:55 No.103969182
>>103969168
>if we applied censorship like China maybe we would be better than China
great take anon, you have more of those?
>If Chinese couldn't buy our houses
it's the jews that buy our houses what are you talking about?
>if we applied censorship like China maybe we would be better than China
great take anon, you have more of those?
>If Chinese couldn't buy our houses
it's the jews that buy our houses what are you talking about?
Anonymous 01/20/25(Mon)12:13:08 No.103969394
Anonymous 01/20/25(Mon)13:28:11 No.103970199