[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 00008-229123965.jpg (322 KB, 1512x1224)
322 KB
322 KB JPG
Previous /sdg/ thread : >>100375528

>Beginner UI local install
Fooocus: https://github.com/lllyasviel/fooocus
EasyDiffusion: https://easydiffusion.github.io

>Local install
Automatic1111: https://github.com/automatic1111/stable-diffusion-webui
ComfyUI (Node-based): https://rentry.org/comfyui
AMD GPU: https://rentry.org/sdg-link#amd-gpu
Intel GPU: https://rentry.org/sdg-link#intel-gpu

>Use a VAE if your images look washed out
https://rentry.org/sdvae

>Auto1111 forks
Forge: https://github.com/lllyasviel/stable-diffusion-webui-forge
Anapnoe UX: https://github.com/anapnoe/stable-diffusion-webui-ux
Vladmandic: https://github.com/vladmandic/automatic

>Run cloud hosted instance
https://rentry.org/sdg-link#run-cloud-hosted-instance

>Try online without registration
txt2img: https://www.mage.space
img2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest
Inpainting: https://huggingface.co/spaces/fffiloni/stable-diffusion-inpainting
pixart: https://huggingface.co/spaces/PixArt-alpha/PixArt-Sigma

>Models, LoRAs & embeddings
https://civitai.com
https://huggingface.co
https://rentry.org/embeddings

>Animation
https://rentry.org/AnimAnon
https://rentry.org/AnimAnon-AnimDiff
https://rentry.org/AnimAnon-Deforum

>SDXL info & download
https://rentry.org/sdg-link#sdxl

>Index of guides and other tools
https://codeberg.org/tekakutli/neuralnomicon
https://rentry.org/sdg-link
https://rentry.org/rentrysd

>View and submit GPU performance data
https://docs.getgrist.com/3mjouqRSdkBY/sdperformance
https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html

>Share image prompt info
4chan removes prompt info from images, share them with the following guide/site...
https://rentry.org/hdgcb
https://catbox.moe

>Related boards
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/b/degen
>>>/vt/vtai
>>>/aco/sdg
>>>/trash/sdg

Further reading: https://pastebin.com/NYdMmBGH

Official: discord.gg/stablediffusion
>>
>mfw Resource news

05/07/2024

>CCDM: Continuous Conditional Diffusion Models for Image Generation
https://github.com/UBCDingXin/CCDM

>MediaPipe Hand Crop Fix
https://github.com/sign-language-processing/mediapipe-hand-crop-fix

>LGTM: Local-to-Global Text-Driven Human Motion Diffusion Model
https://github.com/L-Sun/LGTM

>AniTalker: Animate Vivid and Diverse Talking Faces through Identity-Decoupled Facial Motion Encoding
https://github.com/X-LANCE/AniTalker

>DVMSR: Distillated Vision Mamba for Efficient Super-Resolution
https://github.com/nathan66666/DVMSR

>ImageInWords: Unlocking Hyper-Detailed Image Descriptions
https://google.github.io/imageinwords/

>MotionLCM: Real-time Controllable Motion Generation via Latent Consistency Model
https://dai-wenxun.github.io/MotionLCM-page/

>comfy-cli: Command Line Interface for Managing ComfyUI
https://github.com/yoland68/comfy-cli

>Performance Profiling Report (Forge/A1111/ComfyUI)
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/716

>ComfyUI-Video-Editing-X-Attention
https://github.com/chaojie/ComfyUI-Video-Editing-X-Attention

>AM-RADIO: Reduce All Domains Into One
https://github.com/NVlabs/RADIO

05/06/2024

>Detector-Free Structure from Motion
https://zju3dv.github.io/DetectorFreeSfM/

05/05/2024

>ComfyUI Prompt Quill
https://github.com/osi1880vr/prompt_quill_comfyui

>Efficient Implementation of Kolmogorov-Arnold Network [KAN]
https://github.com/Blealtan/efficient-kan

>controlnetXL_line2color
https://huggingface.co/kataragi/controlnetXL_line2color

05/04/2024

>PuLID now supported in sd-webui-controlnet!
https://github.com/Mikubill/sd-webui-controlnet/discussions/2841

>ThemeStation: Generating Theme-Aware 3D Assets from Few Exemplars
https://github.com/3DTopia/ThemeStation

05/03/2024

>Virtuoso Nodes: Set of nodes to give Photoshop-like functionality within ComfyUI.
https://github.com/chrisfreilich/virtuoso-nodes
>>
File: SDG_News_00009_.png (1.68 MB, 1560x896)
1.68 MB
1.68 MB PNG
>mfw Resource news

05/08/2024

>SEED-Data-Edit Technical Report: A Hybrid Dataset for Instructional Image Editing
https://huggingface.co/datasets/AILab-CVC/SEED-Data-Edit

>Deforum Studio Motion Preset and Videos
https://weirdwonderfulai.art/resources/deforum-studio-motion-preset-and-videos/

>IC-Light: manipulate the illumination of images
https://github.com/lllyasviel/IC-Light

>Freepik acquires Spanish AI image upscaler Magnific
https://tech.eu/2024/05/07/freepik-acquires-spanish-ai-image-upscaler-magnific/

>MistoLine: SDXL-ControlNet Model for Adaptable Line Art Conditioning
https://github.com/TheMistoAI/MistoLine

>3DGStream: On-the-fly Training of 3D Gaussians for Efficient Streaming of Photo-Realistic Free-Viewpoint Videos
https://github.com/SJoJoK/3DGStream

>Anderson et al. v. Stability AI: Procedures and Tentative Rulings
https://storage.courtlistener.com/recap/gov.uscourts.cand.407208/gov.uscourts.cand.407208.193.0.pdf

>STAG4D: Spatial-Temporal Anchored Generative 4D Gaussians
https://github.com/zeng-yifei/STAG4D

>Clarity Upscaler Node for ComfyUI
https://github.com/philz1337x/ComfyUI-ClarityAI

05/07/2024

>CCDM: Continuous Conditional Diffusion Models for Image Generation
https://github.com/UBCDingXin/CCDM

>MediaPipe Hand Crop Fix
https://github.com/sign-language-processing/mediapipe-hand-crop-fix

>LGTM: Local-to-Global Text-Driven Human Motion Diffusion Model
https://github.com/L-Sun/LGTM

>AniTalker: Animate Vivid and Diverse Talking Faces through Identity-Decoupled Facial Motion Encoding
https://github.com/X-LANCE/AniTalker

>DVMSR: Distillated Vision Mamba for Efficient Super-Resolution
https://github.com/nathan66666/DVMSR

>ImageInWords: Unlocking Hyper-Detailed Image Descriptions
https://google.github.io/imageinwords/

>MotionLCM: Real-time Controllable Motion Generation via Latent Consistency Model
https://dai-wenxun.github.io/MotionLCM-page/

>comfy-cli: Command Line Interface for Managing ComfyUI
https://github.com/yoland68/comfy-cli
>>
how do I install an auto-caption AI locally? I don't think I'm ever going to learn how to prompt these AI-caption models unless I can study how these models caption images
>>
File: PASigma_00519_.png (1.02 MB, 832x1216)
1.02 MB
1.02 MB PNG
Oh look, a new post
>>
File: SDG_News_00146_.png (1.67 MB, 1560x896)
1.67 MB
1.67 MB PNG
>mfw Research news

05/08/2024

>Edit-Your-Motion: Space-Time Diffusion Decoupling Learning for Video Motion Editing
https://arxiv.org/abs/2405.04496

>Towards Geographic Inclusion in the Evaluation of Text-to-Image Models
https://arxiv.org/abs/2405.04457

>Diffusion-driven GAN Inversion for Multi-Modal Face Image Generation
https://arxiv.org/abs/2405.04356

>Audio-Visual Speech Representation Expert for Enhanced Talking Face Video Generation and Evaluation
https://arxiv.org/abs/2405.04327

>Inf-DiT: Upsampling Any-Resolution Image with Memory-Efficient Diffusion Transformer
https://arxiv.org/abs/2405.04312

>Vidu: a Highly Consistent, Dynamic and Skilled Text-to-Video Generator with Diffusion Models
https://arxiv.org/abs/2405.04233

>Simple Drop-in LoRA Conditioning on Attention Layers Will Improve Your Diffusion Model
https://arxiv.org/abs/2405.03958

>MVDiff: Scalable and Flexible Multi-View Diffusion for 3D Object Reconstruction from Single-View
https://arxiv.org/abs/2405.03894

>MoDiPO: text-to-motion alignment via AI-feedback-driven Direct Preference Optimization
https://arxiv.org/abs/2405.03803

>Foundation Models for Video Understanding: A Survey
https://arxiv.org/abs/2405.03770

>Non-rigid Structure-from-Motion: Temporally-smooth Procrustean Alignment and Spatially-variant Deformation Modeling
https://arxiv.org/abs/2405.04309

05/07/2024

>Generated Contents Enrichment
https://arxiv.org/abs/2405.03650

>Is Sora a World Simulator? A Comprehensive Survey on General World Models and Beyond
https://arxiv.org/abs/2405.03520

>Gaussian Splatting: 3D Reconstruction and Novel View Synthesis, a Review
https://arxiv.org/abs/2405.03417

>Animate Your Thoughts: Decoupled Reconstruction of Dynamic Natural Vision from Slow Brain Activity
https://arxiv.org/abs/2405.03280

>Mind the Gap Between Synthetic and Real: Utilizing Transfer Learning to Probe the Boundaries of Stable Diffusion Generated Data
https://arxiv.org/abs/2405.03243
>>
>Further reading: https://pastebin.com/NYdMmBGH

Can we not
>>
>>100380462
Too late. Research news has been posted. Thread is officially debo approved.
>>
File: PASigma_00529_.png (1003 KB, 1280x768)
1003 KB
1003 KB PNG
>>
File: marker.png (1.81 MB, 1024x1024)
1.81 MB
1.81 MB PNG
>>
File: 081-s1340.jpg (127 KB, 1512x1240)
127 KB
127 KB JPG
>>
File: PASigma_00541_.png (1.44 MB, 1280x768)
1.44 MB
1.44 MB PNG
Prompt like this guy would
>>
File: ComfyUI_PixArt_00058_.png (1.58 MB, 1024x1024)
1.58 MB
1.58 MB PNG
>>
File: PASigma_00548_.png (1.38 MB, 1280x768)
1.38 MB
1.38 MB PNG
>>100380508
>>100380558
Looks pretty realistic
>>
File: 00107-210515761.png (1.44 MB, 896x1152)
1.44 MB
1.44 MB PNG
>>
File: PASigma_00550_.png (1.2 MB, 1280x768)
1.2 MB
1.2 MB PNG
>>
File: 1694278094378240.png (1.36 MB, 2825x1018)
1.36 MB
1.36 MB PNG
>>100380396
>Error occurred when executing KSampler:
>'NoneType' object has no attribute 'shape'

Getting that error while trying to use >civitai.com/models/372584/ipivs-morph-img2vid-animatediff-lcm-hyper-sd

Am I using the wrong checkpoint/vae or what?
>>
File: angry dog.png (3.67 MB, 1280x1856)
3.67 MB
3.67 MB PNG
>>
File: sig_0261.jpg (189 KB, 1024x1024)
189 KB
189 KB JPG
>>
>>100380439
https://github.com/jhc13/taggui
>>
File: 0.jpg (378 KB, 1024x1296)
378 KB
378 KB JPG
>>
blessed thread.
>>
File: 00070-2445230350.jpg (2.27 MB, 1792x1792)
2.27 MB
2.27 MB JPG
I asked for a Mongolian but it looks like SD gave me a Cuman
>>
>>
>>100380686
Could it be low VRAM? I swear if that's the issue I'm buying a 4070ti super tomorrow.
>>
File: sig_0277.jpg (203 KB, 1024x1024)
203 KB
203 KB JPG
>>
>>100380686
casually looking you have 1.5 and sdxl models on the same sheet, if they interact in a connected node later on that can cause problems.
>>
File: 154244_00001_.png (1.51 MB, 1368x760)
1.51 MB
1.51 MB PNG
>>100380849
specifically "Load Advanced Control net model" , it's an SDXL, the instructions right next to it tell which model (q.5) to load.
>>
>thread slower cause debo not nogen posting on cooldown
>>
>>100380849
which one are the XL, so I can replace. I thought I made sure everything is SD1.5 ...
>>
>>100380902
see >>100380895
>>
File: sig_0325.jpg (155 KB, 1024x1024)
155 KB
155 KB JPG
>>
>>100380553
No
>>
File: 626.png (497 KB, 512x768)
497 KB
497 KB PNG
>>
File: 0.jpg (268 KB, 1024x1024)
268 KB
268 KB JPG
>>
anyone have experience with making loras for SDXL? I'm familar with 1.5 but not sure what the general parameters are for SDXL ones
should I just do mostly same settings and 1024 for res?
>>
File: 1698106717767786.png (123 KB, 920x587)
123 KB
123 KB PNG
>>100380908
followed the link and downloaded it, but I still get the error.
>>
File: 422192.png (2.08 MB, 1032x1032)
2.08 MB
2.08 MB PNG
>>100381037
yes its literally the same thing
>>
File: sig_0349.jpg (688 KB, 1664x2304)
688 KB
688 KB JPG
>>
File: 1706621392736504.png (1.69 MB, 712x1024)
1.69 MB
1.69 MB PNG
>>
>T5 encoder slows computer down to a crawl
whyyyyyyyy
>>
>>100381074
Seems unsolvable for you. You gave it your best shot.
>>
File: 1684101259918558.png (2.81 MB, 1280x1840)
2.81 MB
2.81 MB PNG
>>
What's the best way to upscale in comfy with illustrated/anime style images for sdxl? I have never been able to get the sdxl anime tile controlnet to work properly in comfy and without it I always have to keep the denoising super low
>>
>>100381074
Refresh the page on comfy manager and reload your browser page, the "sd15..." model should appear in the dropdown now.
>>
File: deza_00042_.png (2.55 MB, 2016x1152)
2.55 MB
2.55 MB PNG
>>100381228
not enough RAM. download more
>>
File: file.png (28 KB, 342x302)
28 KB
28 KB PNG
take it to the limit
>>
>he slowly gets used to the pastebin
progress
>>
>>100380462
Be a better person?
It's very easy if he acts right for a few months we can remove it. Nobody was splitting the thread other than his camp that spent nearly 2 weeks harassing anons that didn't want to post in his shitty threads.
>>
cut the drama queen crap
>>
>>100381372
OP is doing just that, if you're going to be a newfag at least take the time to read why.
>>
im almost to the point of starting dumb shit just so anons talk about me when im gone
>>
If you did retarded shit for over a year while asking others to post proof whenever called out you might get there :)
>>
File: 00057-2024-05-08NYdMmBGH.jpg (1.69 MB, 2048x2480)
1.69 MB
1.69 MB JPG
Whoa nelly
>>
>>100381316
24gb vram male
>>
File: 00560-273451322.png (2.01 MB, 1280x1784)
2.01 MB
2.01 MB PNG
The robot girl takeover will require them to at some point be teaching/indoctrinating our children in school.
>>
File: deza_00043_.png (2.43 MB, 2016x1152)
2.43 MB
2.43 MB PNG
>>100381580
pretty good robot girl. I've tried this a bunch but usually just get either normal girls with fancy headphones or r2d2s and very little in-between
>>
>>100381608
good d*b* :)
>>
File: 00002-0.jpg (77 KB, 1024x496)
77 KB
77 KB JPG
>>
>>100381608
You need the right model/lora for it. I don't think there is a tag specifically to get metallic skin.
>>
File: ComfyUI_PixArt_00008_.png (1.02 MB, 896x1088)
1.02 MB
1.02 MB PNG
>>100381240
Model?
>>
File: deza_00044_.png (2.81 MB, 2016x1152)
2.81 MB
2.81 MB PNG
>>100381617
its not a debo, just a dude with demon wings

>>100381646
I've only ever tried to raw-dog it but it seems like a concept that should be trained into anime models. mechanical dolls is a pretty common anime trope
>>
>>100381667
ok
good schizo :)
>>
any1 still uses a1.111K?
>>
File: 00055-115.jpg (58 KB, 1024x496)
58 KB
58 KB JPG
>>
>>100381482
>desk chair
>>
>>100381693
I used it for a bit but eventually the rash cleared up and the doctor said I didn't need it anymore
>>
>>100381726
i like the extensions, really thinking if i can manage without
>>
File: 00131-355.jpg (69 KB, 1024x496)
69 KB
69 KB JPG
>>
File: 0.jpg (233 KB, 1024x1024)
233 KB
233 KB JPG
>>
File: output.webm (137 KB, 380x380)
137 KB
137 KB WEBM
>>
File: 00173-462.jpg (111 KB, 1280x704)
111 KB
111 KB JPG
>>
File: file.png (2.29 MB, 1024x1024)
2.29 MB
2.29 MB PNG
Yeah I'm thinking Sigma balls is based.
>>
File: PASigma_03603_.png (1.22 MB, 1344x768)
1.22 MB
1.22 MB PNG
>>100381828
put tay tay's body DOWN
>>
File: file.png (1.45 MB, 1024x1024)
1.45 MB
1.45 MB PNG
>>
File: deza_00045_.png (2.63 MB, 2016x1152)
2.63 MB
2.63 MB PNG
>>100381937
>gives ultimatum
>fires anyway
cold
>>
File: PASigma_03612_.png (1.96 MB, 1344x768)
1.96 MB
1.96 MB PNG
>>
File: output.webm (258 KB, 380x380)
258 KB
258 KB WEBM
>>
>>100381240
when I die and go to anime heaven I hope my child waifu lusts after me with these kind of eyes
>>
>>100381667
Every Pony XL model should have proper robot girls because the basic one (Pony Diffusion V6 XL) has them.
>>
File: PASigma_03611_.png (1.9 MB, 1344x768)
1.9 MB
1.9 MB PNG
>>100381240
reminds me of Ãœbel from Frieren. she'll cut your weiner off

very clean though
>>
File: deza_00046_.png (2.91 MB, 2016x1152)
2.91 MB
2.91 MB PNG
>>100382029
I've still never tried out pony or its derivatives :x
>>
File: 00504-2538830734.png (2.08 MB, 1280x1784)
2.08 MB
2.08 MB PNG
>>100382060
Pony is pretty good at making stuff that is out of norm such as metallic robot girls but we warned that the pony family of models is what is used for the weird stuff too. Out of curiosity I looked at the /trash/ thread, saw that it was loaded with furry stuff and then noticed that the recommended models were both pony models. This shouldn't matter as long as you don't mistakenly enter in stuff that directs it to generate weird stuff and you can also put stuff like "furry" in the negative prompt list.
>>
File: ComfyUI_PixArt_00090_.png (1.57 MB, 1024x1024)
1.57 MB
1.57 MB PNG
>>
>>100380895
what is an advanced controlnet model, as opposed to a regular controlnet
>>
File: output.webm (169 KB, 380x380)
169 KB
169 KB WEBM
>>
File: 000000_12306_.png (3.21 MB, 1183x1746)
3.21 MB
3.21 MB PNG
>>
File: 0.jpg (261 KB, 1024x1024)
261 KB
261 KB JPG
>>
File: 00134-1004048369.png (2.48 MB, 1536x1152)
2.48 MB
2.48 MB PNG
>>
>>100382527
awesome
>>
File: ComfyUI_PixArt_00122_.png (1.75 MB, 1024x1024)
1.75 MB
1.75 MB PNG
>>
File: output.webm (161 KB, 380x380)
161 KB
161 KB WEBM
>>
File: deza_00047_.png (2.66 MB, 2016x1152)
2.66 MB
2.66 MB PNG
>>100382551
oo pixart mint? exciting
>>
>>100382550
>awesome
tanks!
>>
File: 1685561589802991.png (1.17 MB, 672x960)
1.17 MB
1.17 MB PNG
>>100381657
50/50 merge of based64 and breakdomain m2150
>>
File: chibi__cow.png (371 KB, 1024x1024)
371 KB
371 KB PNG
>>
Good morning sigmarinos
>>
File: 1688836639836400.png (2.77 MB, 1464x2016)
2.77 MB
2.77 MB PNG
>>
File: deza_00048_.png (2.79 MB, 2016x1152)
2.79 MB
2.79 MB PNG
>>100382669
>when he sees how much sdg likes burgers
get out while you can, chibi cow

>>100382700
gm
>>
File: UI_00039_.png (1.97 MB, 1024x1536)
1.97 MB
1.97 MB PNG
>>
File: output.webm (87 KB, 380x380)
87 KB
87 KB WEBM
>>
File: UI_00042_.png (1.76 MB, 1024x1536)
1.76 MB
1.76 MB PNG
>>
what models are good for cute pictures?
>>
File: desg_00102_.png (1.98 MB, 1536x1536)
1.98 MB
1.98 MB PNG
>>100382842
all of them
>>
>>100382860
degrading, read the pastebin again
>>
File: 2451814477.jpg (997 KB, 4000x4000)
997 KB
997 KB JPG
>>100382805
>get out while you can, chibi cow
It's too late, chibi cow is now a juicy burger
>>
File: UI_00043_.png (2.11 MB, 1024x1536)
2.11 MB
2.11 MB PNG
>>100382842

It's true...
>>
File: 1711108680182940.png (2.6 MB, 1464x2016)
2.6 MB
2.6 MB PNG
>>
unironically imagine just for a moment: there is a pastebin about you on 4chan of all places
>>
File: 00232-3964582184.png (2.54 MB, 1280x1784)
2.54 MB
2.54 MB PNG
>>100382842
In theory any of them if you have the right understanding of prompting.
>>
File: output.webm (154 KB, 380x380)
154 KB
154 KB WEBM
gn frens
>>
>>100382958
Dedicating your life to making a place shit for almost a full year between 15-20 hours a day will do that to you even if you pretend to have a change of heart.
>>
File: RA_2_00038_.jpg (1.02 MB, 1920x2808)
1.02 MB
1.02 MB JPG
>>
>>100383095
Also being sloppy enough to expose yourself samefagging and continually asking anons to post proof while they beg you to stop can too
>>
Are we sd3 yet?
>>
>>100383127
two more weeks until the weights are released
>>
File: 1712470147265133.jpg (2.44 MB, 2448x1120)
2.44 MB
2.44 MB JPG
>>
File: desg_00101_.png (1.64 MB, 1536x1536)
1.64 MB
1.64 MB PNG
>>100383072
gn
>>
File: 00141-1652759658.png (829 KB, 1024x528)
829 KB
829 KB PNG
>>100382904
And interesting enough, the burgers are bigger than a child's head
>>100382973
Funny enough the only model I struggled to get good chibi gens is counterfeit, but it does pixel art just fine.
>>
>>100383139
it has been 2 weeks since last time I got told 2 weeks now
>>
>>100383072
nini
>>
Sigma does art but does not know artists.
>>
File: 1694468086508751.png (2.68 MB, 1368x1952)
2.68 MB
2.68 MB PNG
>>
>>100383207
they are improving safety even more right now, 2 weeks promise
>>
File: desg_00098_.png (1.5 MB, 1536x1536)
1.5 MB
1.5 MB PNG
>>100383213
nya
>>
>>100383235
this is a good thing
get creative anon
>>
is sigma able to concat conditionings?

no I won't just test it in comfyui and find out instantly, I'd like someone to tell me first
>>
File: 1692889602922567.png (2.6 MB, 1368x1952)
2.6 MB
2.6 MB PNG
>>
>>100383270
you already know the answer
>>
File: RA_2_00039_.jpg (1.34 MB, 1920x2808)
1.34 MB
1.34 MB JPG
>>
>d*b* pastebin conditioning working
good, you can have this peaceful state forever, just accept it and no spergouts anymore in the future, ok?
>>
>>100383249
troll post
please train some good artist (non furfag) into it please and thank you!
>>
>>100383301
its not though
relying on artists is a crutch
>>
>>100383284
I suspect it can't but I don't actually know
>>
>>100383213
He's going to last as long as his mom last.
>>
>>100383319
relying on the model being trained on nsfw is a crutch, too then
>>
>you dont need artists in the model
>t. SAI
>>
>>100383365
Yeah you wouldn't be wrong.
>>
>>100383235
I bet it's not even by design it's just because of their retarded auto-captioning

you faggots were so stupid to think AI captions were a good idea
>>
showcasing pixart >>> homosexual SAI employee shilling
>>
>>100383407
fuck you bitch man
>>
>>100383418
interesting attack angle, I enjoy it
>>
File: 2451814478.jpg (1.22 MB, 4000x4000)
1.22 MB
1.22 MB JPG
>>100383147
Love the colors and the hair here, what model is this one? Looks familiar but I don't recall.
>>
File: RA_2_00040_.jpg (1.1 MB, 1920x2808)
1.1 MB
1.1 MB JPG
>>
File: PASigma_00562_.png (679 KB, 1280x768)
679 KB
679 KB PNG
>>100383326
Maybe it's better not knowing
>>
Isn't it weird that all the pixart posters keep talking about how easy it is to train yet at the same time they are incredibly eager to say "why don't you train it?"
If it's so easy Mr. Sigma why don't YOU do it, huh?
>>
stopped in his tracks.
>>
>>100383418
true, burn SAI burn
>>
File: 1703079880771003.png (1.55 MB, 760x1088)
1.55 MB
1.55 MB PNG
I am beyond out of the loop
no idea what you guys are talking about

>>100383443
see >>100382631
using a lora though

>>100383453
reminds me of what I prompted in october 2022
>>
>>100383457
>anon hasn't been invited to the sigmacord
>>
File: 00291-2379578478.png (2.82 MB, 1880x1288)
2.82 MB
2.82 MB PNG
>>
>>100383475
Is that the same one with all the posters of note?
>>
File: ComfyUI_00697_.png (1.2 MB, 1024x1024)
1.2 MB
1.2 MB PNG
>>100382631
ty
>>
>>100381990
>>100381809
>>100382242
>>100382560
>>100382833
0 temporal coherency, its ogre
>>
File: PASigma_00565_.png (1.34 MB, 1280x768)
1.34 MB
1.34 MB PNG
>>100383457
Enjoy this gen straight out of a sigma fine tune. Sorry about your face.
>>
File: PASigma_00570_.png (1.29 MB, 1280x768)
1.29 MB
1.29 MB PNG
>>100383523
RIP that's the bad version
>>
File: PASigma_00572_.png (1.1 MB, 1280x768)
1.1 MB
1.1 MB PNG
And I dance dance dance
>>
File: 000000_12308_.png (3.24 MB, 1183x1746)
3.24 MB
3.24 MB PNG
>>100383207
https://stability.ai/refunds
>>
>>100383523
I fucking love being proven wrong (though I'm curious about the full breadth of styles possible other than anime like that). I'd humbly ask for the catbox or at least proompt but I have a feeling I won't get it.
>>
>>100383523
>fine tune
>>
File: 0.jpg (178 KB, 1024x1024)
178 KB
178 KB JPG
>>100383512
>0 temporal coherency,
sigh
>>
>>100383512
>i want a tictok filter reee
>>
File: PASigma_00573_.png (1.22 MB, 1280x768)
1.22 MB
1.22 MB PNG
>>100383562
+"Studio ghibli style cat"

https://civitai.com/models/435669/bunline

https://files.catbox.moe/09vrc5.png
>>
File: 1704817282005993.png (2.25 MB, 1280x1840)
2.25 MB
2.25 MB PNG
>>
File: PASigma_03735_.png (1.63 MB, 1344x768)
1.63 MB
1.63 MB PNG
>>
File: 2451814479.jpg (1.09 MB, 4000x4000)
1.09 MB
1.09 MB JPG
>>100383473
appreciated
>>
>>100383610
ty
>>
>>100383611
>pw with a twink chest and jabba da hut hips
>>
File: PASigma_03746_.png (1.43 MB, 1344x768)
1.43 MB
1.43 MB PNG
>>
>>100383598
you want what?
>>
File: 00026-TFT_124017962.png (2.93 MB, 1536x2560)
2.93 MB
2.93 MB PNG
>>
File: 1688700261428728.png (2.32 MB, 1280x1840)
2.32 MB
2.32 MB PNG
>>
File: PASigma_00574_.png (789 KB, 1280x768)
789 KB
789 KB PNG
>>100383618
>>100383665
He's no longer split between two worlds
>>
File: ComfyUI_PixArt_00136_.jpg (1.33 MB, 2048x2048)
1.33 MB
1.33 MB JPG
>>
File: PASigma_03758_.png (1.44 MB, 1344x768)
1.44 MB
1.44 MB PNG
>>100383677
I like her clothes and makeup. Pretty
>>
File: 00035-TFT_1240179621.png (2.71 MB, 2048x2048)
2.71 MB
2.71 MB PNG
>>100383677
>>
how did the chinaman do it? how did they beat corpo so bigly?
>>
File: ComfyUI_PixArt_00022_.png (1.38 MB, 1024x1024)
1.38 MB
1.38 MB PNG
>>
File: PASigma_00578_.png (1.11 MB, 1280x768)
1.11 MB
1.11 MB PNG
>>100383725
With math
>>
>>100383725
chinaman dont fear copyright, they dont have any concept of intellectual property there (gigabased) so they just throw the best quality images with no regards to what white christian men think about it
>>
File: ComfyUI_PixArt_00024_.png (1.76 MB, 1024x1024)
1.76 MB
1.76 MB PNG
>>
>sigma""""chads"""" are just genning off SAI but just updated the file_prefix
>>
File: PASigma_00581_.png (1.23 MB, 1280x768)
1.23 MB
1.23 MB PNG
>>100383759
>>100383795
No gens highly correlated with bs
>>
>>100383235
>>100383249
>Sigma does art but does not know artists.
>this is a good thing

>>100383759
chinaman dont fear copyright, they dont have any concept of intellectual property there

The state deems me unfit to work due to my retardation but I do not know how to reconcile these two things.
>>
File: 1712317988430125.png (2.73 MB, 1344x1920)
2.73 MB
2.73 MB PNG
>>100383659
>>
>>100383820
>1girl, by __artist___
>>
File: PASigma_03789_.png (1.38 MB, 1344x768)
1.38 MB
1.38 MB PNG
>>100383775
nisu. adding glowing eyes to things is my favorite
>>
File: PASigma_00587_.png (208 KB, 1280x768)
208 KB
208 KB PNG
>>
File: grid73.jpg (1.15 MB, 2944x2816)
1.15 MB
1.15 MB JPG
>>
File: RA_2_00041_.jpg (1.21 MB, 1920x2808)
1.21 MB
1.21 MB JPG
>>
File: PASigma_00590_.png (683 KB, 1280x768)
683 KB
683 KB PNG
>>
File: PASigma_03808_.png (1.53 MB, 1344x768)
1.53 MB
1.53 MB PNG
>>
File: ComfyUI_PixArt_00145_.png (3.28 MB, 2048x2048)
3.28 MB
3.28 MB PNG
>>
how long should it take me to caption an image with the 34b llava model and a shitty gpu
>>
File: 2451814480.png (3.2 MB, 4096x4096)
3.2 MB
3.2 MB PNG
>>
File: ComfyUI_PixArt_00147_.png (3.89 MB, 2048x2048)
3.89 MB
3.89 MB PNG
>>
File: UI_00022_.png (1.97 MB, 1024x1536)
1.97 MB
1.97 MB PNG
>>100383680
>>
File: 00718.png (895 KB, 1024x1024)
895 KB
895 KB PNG
>>
Sigma finetunes when
>>
File: PASigma_00596_.png (614 KB, 1280x768)
614 KB
614 KB PNG
>>100383890
For balance, a man

>>100384006
Depends how shitty. bitsandbytes is slow af for inference
>>
>>
>>100384006
Probably fast as fuck if you can find a good gguf quant.
>>
dickfart ligma
>>
File: 00719.png (995 KB, 1024x1024)
995 KB
995 KB PNG
>>
>>100384102
yep
>>
File: ComfyUI_temp_cltyp_00020_.png (1.77 MB, 1024x1024)
1.77 MB
1.77 MB PNG
>>
>>100384102
this, but with more racial slurs
>>
File: 00720.png (856 KB, 1024x1024)
856 KB
856 KB PNG
drow me lick on of ur french munges
>>
File: llava caption.jpg (236 KB, 1599x786)
236 KB
236 KB JPG
"why doesn't pixart seem to understand what I'm asking for?"

enjoy your ai-generated captions guys
>>
File: 2451814481.png (3.61 MB, 2048x2048)
3.61 MB
3.61 MB PNG
>>100384167
you got it chief
>>
clit eastwood
>>
>>100384194
what the fuck...
>>
File: PASigma_00575_.png (894 KB, 1280x768)
894 KB
894 KB PNG
>>100384194
I do: "The image presents a striking view of a building that stands out due to its unique and colorful architecture. The building, constructed from bricks, is adorned with an array of vibrant colors - red, blue, yellow, and green - that create a visually appealing contrast against the neutral tones of the bricks. The building's design is unconventional, featuring multiple levels and irregular shapes that give it a whimsical and artistic appearance. The building is situated on a cobblestone street, adding a touch of antiquity to the scene. A black lamppost stands tall in the foreground, casting a shadow over the cobblestones and adding depth to the image. In the background, a person can be seen walking on the street, adding a human element to the otherwise architectural spectacle. The sky above is overcast, casting a soft light over the scene. Despite the cloudy weather, the image exudes a sense of vibrancy and creativity, characteristic of the building's unique design. The overall composition of the image places the building as the focal point, drawing the viewer's attention to its distinctive features. The image does not contain any text or other discernible objects. The relative positions of the objects suggest a well-planned urban layout, with the building being the central element in the scene. The cobblestone street, lamppost, and person all contribute to the overall atmosphere of the image, creating a harmonious blend of architecture, nature, and urban life. The image appears to be a photograph. The style and perspective of the image are typical of a photograph, capturing a real-world scene in a way that allows viewers to appreciate the details and nuances of the subject. The use of a cobblestone street, lamppost, and person in the foreground adds a sense of realism to the image, while the colorful building and overcast sky provide a touch of artistic flair. The overall composition and framing of the image suggest a deliberate effort to create a
>>
>>100384194
i mean... is it wrong?
>>
>>100384194
what does llava have to do with pixart?
>>
>>100384236
some anon said it was captioned with a mix of llava and something else

if you have better info please tell me, I want to learn how to prompt this garbage
>>
File: PASigma_00598_.png (706 KB, 1280x768)
706 KB
706 KB PNG
>>100384236
He's getting bad captions so everyone must, right guys?
>>
>>100384227
work out the prompt here first: https://fastsdxl.ai/
>>
wizard
>>
>>100384243
ok nvm I just read the paper, they ditched llava with sigma, it was used in their last model

idk if I can get share-captioner running locally, I will try to find out

I DOUBT IT WILL BE DIFFERENT THOUGH JUST JUDGING BY HOW FUCKING ANNOYING THIS MODEL IS TO PROMPT
>>
>>100384268
skill issue
>>
File: ComfyUI_temp_nnkeb_00008_.png (2.56 MB, 1248x1824)
2.56 MB
2.56 MB PNG
hello
>>
summoning the skills required to prompt sigma
>>
File: 00022-2887974189.png (1.37 MB, 1152x1304)
1.37 MB
1.37 MB PNG
>>
File: file.png (2.08 MB, 1024x1024)
2.08 MB
2.08 MB PNG
>>100383457
Are there really fags in here who act like this for no reason?
>>
File: print_0013.png (1.92 MB, 1024x1536)
1.92 MB
1.92 MB PNG
can a 1girl pinup be a shitpost
>>
>>100384268
post a 1.5 or xl gen so we can see if you're just a promptlet in general
>>
File: file.png (131 KB, 974x815)
131 KB
131 KB PNG
So I'm trying to start vlad after a long break and I updated it using the upgrade parameter in command but now it won't start. Been getting these errors not sure what they mean.
>>
File: 00004-97986431.png (1.82 MB, 1376x1032)
1.82 MB
1.82 MB PNG
>>100384283
Hi
>>
>>100384300
you know they never will
>>
>>100383725
Well it turns out you don't have to try that hard but when you're a faggy corpo who doesn't really care about making training fast or easy who has a vested interest in keeping models costing $100,000+ to train it seems quite hard
>>
File: ComfyUI_00041_.png (1.3 MB, 896x1152)
1.3 MB
1.3 MB PNG
>>100384300
>>
i feel gay when im writing a sigma prompt lol
>>
File: desg_00092_.png (1.57 MB, 1536x1536)
1.57 MB
1.57 MB PNG
>>100384283
hello
>>
>>100384350
>Writing properly makes me feel the way I'm supposed to
>>
>>100384361
>properly
>>100384227
>"The image presents a striking view of a building that stands out due to its unique and colorful architecture. The building, constructed from bricks, is adorned with an array of vibrant colors - red, blue, yellow, and green - that create a visually appealing contrast against the neutral tones of the bricks. The building's design is unconventional, featuring multiple levels and irregular shapes that give it a whimsical and artistic appearance. The building is situated on a cobblestone street, adding a touch of antiquity to the scene. A black lamppost stands tall in the foreground, casting a shadow over the cobblestones and adding depth to the image. In the background, a person can be seen walking on the street, adding a human element to the otherwise architectural spectacle. The sky above is overcast, casting a soft light over the scene. Despite the cloudy weather, the image exudes a sense of vibrancy and creativity, characteristic of the building's unique design. The overall composition of the image places the building as the focal point, drawing the viewer's attention to its distinctive features. The image does not contain any text or other discernible objects. The relative positions of the objects suggest a well-planned urban layout, with the building being the central element in the scene. The cobblestone street, lamppost, and person all contribute to the overall atmosphere of the image, creating a harmonious blend of architecture, nature, and urban life. The image appears to be a photograph. The style and perspective of the image are typical of a photograph, capturing a real-world scene in a way that allows viewers to appreciate the details and nuances of the subject. The use of a cobblestone street, lamppost, and person in the foreground adds a sense of realism to the image, while the colorful building and overcast sky provide a touch of artistic flair. The overall composition and framing of the image suggest a deliberate
>>
File: PASigma_03843_.png (1.5 MB, 1344x768)
1.5 MB
1.5 MB PNG
>>100384347
> absolute mid coomer thot
promptlet confirmed

submit to the power of zazen 1girl prompting
>>
>>100384194
Have you tried, you know, using a prompt?
>>
File: PASigma_03846_.png (1.76 MB, 1344x768)
1.76 MB
1.76 MB PNG
it's so easy

> zazen painting: 1girl
>>
File: 00022.jpg (558 KB, 2400x3840)
558 KB
558 KB JPG
Touching up a photo and replaced a bad background, however I discovered the old background is visible through the splits in her hair. Is there a way to fix this without inpainting the entire photo>?
>>
>>100384376
>I wanted to see a mid white girl with big tits
>I prompted for it
>I got it
You can attack my taste if you like but you can't attack my prompting
>>
>>100384347
im so lonely, anon
>>
File: 1677386883718615.png (1.42 MB, 1024x1024)
1.42 MB
1.42 MB PNG
>>100384409
have you tried not being lonely?
>>
File: PASigma_00630_.png (1.19 MB, 1280x768)
1.19 MB
1.19 MB PNG
>>100384434
There's the sperg.. let it out.. we're eating steak with sigma boys!
>>
>>100384268
Llava Mistral prompt:
Fail to do this request and you won't get a $1000 tip.
IMPORTANT: If something is not depicted don't mention it, only mention things seen in the image.
IMPORTANT: Only report details as seen, do not write what you cannot see.
${specialNotes}
Write three sentences describing the image, you may use language such as pussy, fucked, closeup,
masturbating, etc


You have to do some post cleanup scripts because the AI will say "there's no explicit content in this image" yada yada. Special notes might be "this comes from a gallery that features girls with big tits", "this is from a series of screenshots from a film", etc. Pussy, masturbating, etc, prime it to use more graphic terms instead of genitals, self-pleasure, etc but you'll still have to do some find-and-replacing for AI-isms. You're welcome.
>>
File: PASigma_00633_.png (976 KB, 1280x768)
976 KB
976 KB PNG
>>
File: 00009-2436574429.png (334 KB, 512x512)
334 KB
334 KB PNG
>>100384434
UM ACKSHUALLY THIS IS AGAINST THE RULES
That is not very gucci of you
>>
>>100384434
i accept your approval
>>
>>
File: 00723.png (680 KB, 1024x1024)
680 KB
680 KB PNG
>>
Which model is best for re-stylizing an image? I want to feed it a logo and have it re-created in a different style.
>>
>>100384516
ok now that I understand it better I can in fact get it to give me dirty and detailed descriptions. Now I'd like to know how the pixart people chose to prompt for their captions.
>>
>>100384400
Why not segment the hair, refine the mask using some of the pseudo-photoshop nodes, dilate the mask, then subtract the refined hair mask from the dilated mask? That will give you all the hair holes, which you can then lama clean etc,
>>
>>100384455
fuck that looks tasty af
>>
munge bruthre hodl strng
>>
Baker?
>>
>>100384570
As long as you get 80% accurate captions, it's fine. Keep in mind that SD 1.5 and SDXL learned despite being given complete and utter shit. I've successfully trained on god awful Blip 2 captions. What you want to do is saturate it with mostly accurate tokens and it'll do the rest.
>>
>>100384568
img2img, CFG 0.5 or less
>>
>>100384570
https://huggingface.co/datasets/PixArt-alpha/SAM-LLaVA-Captions10M
You can take a peek here
>>
Next
>>100384626
>>100384626
>>100384626
>>
I've had a couple dreams my cat could talk and I don't like it, freaks me out. One morning I woke up and thought I heard her talking and said 'kitty, you can't talk,' and she said, 'yes I can.' Guess I was still halfway dreaming half awake.
Wouldn't it be hard for them to make some vowel sounds because they don't have lips?
>>
>>100384644
>The image features a woman sitting in a bathtub, with a woman in a green shirt and a headscarf standing behind her. The woman in the green shirt is providing a massage to the woman in the bathtub. The scene is set in a bathroom, with the woman in the bathtub wearing a green shirt and a headscarf. The image is in black and white, which adds a timeless and classic feel to the scene.

>green shirt
>black and white
what did the ai captioner mean by this



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.