[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]


Classic prompts new mods edition

Previously on /sdg/: >>101708633

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

>Local install
Automatic1111: https://github.com/automatic1111/stable-diffusion-webui
ComfyUI: https://github.com/comfyanonymous/ComfyUI
SD.Next: https://github.com/vladmandic/automatic
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

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

>Try online without registration
sd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-medium
txt2img: https://www.mage.space
img2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest

>Models, LoRAs & upscaling
https://civitai.com
https://huggingface.co
https://aitracker.art
https://openmodeldb.info

>Black Forest Labs: Flux
https://huggingface.co/black-forest-labs/FLUX.1-schnell
https://huggingface.co/spaces/black-forest-labs/FLUX.1-schnell
https://comfyanonymous.github.io/ComfyUI_examples/flux

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

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

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

>Discord
6wUwtcJsr2

>Related boards
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/b/degen
>>>/vt/vtai
>>>/aco/sdg
>>>/trash/sdg
>>
File: FDG_News_00007_.jpg (930 KB, 1344x960)
930 KB
930 KB JPG
>mfw Resource news

08/03/2024

>ComfyUI/Forge Implementation of Smoothed Energy Guidance
https://github.com/pamparamm/sd-perturbed-attention

>TryOnDiffusion: A Tale of Two UNets
https://github.com/fashn-AI/tryondiffusion

>Nvidia reportedly delays its next AI chip due to a design flaw
https://www.theverge.com/2024/8/3/24212518

>ComfyUI Frontend Modernization: Transitioning to a New Era on August 15, 2024
https://github.com/comfyanonymous/ComfyUI/issues/4169

>CEO of Invoke says Flux fine tunes are not going to happen
https://www.reddit.com/r/StableDiffusion/comments/1eiuxps

>ComfyUI-FLUX-fal-API
https://github.com/gokayfem/ComfyUI-FLUX-fal-API

08/02/2024

>Optimizing Diffusion Models for Joint Trajectory Prediction and Controllable Generation
https://yixiaowang7.github.io/OptTrajDiff_Page

>UniTalker: Scaling up Audio-Driven 3D Facial Animation through A Unified Model
https://github.com/X-niper/UniTalker

>Smoothed Energy Guidance for SDXL
https://github.com/SusungHong/SEG-SDXL

>Mitigating Multilingual Hallucination in Large Vision-Language Models
https://github.com/ssmisya/MHR

>GalleryGPT: Analyzing Paintings with Large Multimodal Models
https://github.com/steven640pixel/GalleryGPT

>The Manga Whisperer: Automatically Generating Transcriptions for Comics
https://github.com/ragavsachdeva/magi

08/01/2024

>Stable Fast 3D: Rapid 3D Asset Generation From Single Images
https://stability.ai/news/introducing-stable-fast-3d

>Announcing Black Forest Labs
https://blackforestlabs.ai/announcing-black-forest-labs

>Flux: The Next Leap in T2I Models
https://blog.fal.ai/flux-the-largest-open-sourced-text2img-model-now-available-on-fal

>ComfyUI: Basic Flux Schnell and Dev implementation
https://github.com/comfyanonymous/ComfyUI/commit/1589b5

>Kolors ipadapter FaceID Plus
https://github.com/Kwai-Kolors/Kolors/tree/master/ipadapter_FaceID

>The EU’s AI Act is now in force
https://techcrunch.com/2024/08/01/the-eus-ai-act-is-now-in-force
>>
File: 0.jpg (326 KB, 1056x1024)
326 KB
326 KB JPG
>>
File: 2697860287.png (1.94 MB, 1144x1280)
1.94 MB
1.94 MB PNG
The Big Apple: where people never dance.
>>
>mfw Research news

08/03/2024

>Image Super-Resolution with Taylor Expansion Approximation and Large Field Reception
https://arxiv.org/abs/2408.00470

>Localized Gaussian Splatting Editing with Contextual Awareness
https://arxiv.org/abs/2408.00083

>Hierarchical Conditioning of Diffusion Models Using Tree-of-Life for Studying Species Evolution
https://arxiv.org/abs/2408.00160

>SynthVLM: High-Efficiency and High-Quality Synthetic Data for Vision Language Models
https://arxiv.org/abs/2407.20756

>Real Face Video Animation Platform
https://arxiv.org/abs/2407.18955

>ObjectCarver: Semi-automatic segmentation, reconstruction and separation of 3D objects
https://arxiv.org/abs/2407.19108

>Multi-Expert Adaptive Selection: Task-Balancing for All-in-One Image Restoration
https://arxiv.org/abs/2407.19139

>Exploring the Adversarial Robustness of CLIP for AI-generated Image Detection
https://arxiv.org/abs/2407.19553

>Advancing Prompt Learning through an External Layer
https://arxiv.org/abs/2407.19674

>VolDoGer: LLM-assisted Datasets for Domain Generalization in Vision-Language Tasks
https://arxiv.org/abs/2407.19795

>Mixture of Nested Experts: Adaptive Processing of Visual Tokens
https://arxiv.org/abs/2407.19985

>Perm: A Parametric Representation for Multi-Style 3D Hair Modeling
https://cs.yale.edu/homes/che/projects/perm/

>ML-Mamba: Efficient Multi-Modal Large Language Model Utilizing Mamba-2
https://arxiv.org/abs/2407.19832

>Adversarial Robustness in RGB-Skeleton Action Recognition: Leveraging Attention Modality Reweighter
https://arxiv.org/abs/2407.19981

>Exploring Robust Face-Voice Matching in Multilingual Environments
https://arxiv.org/abs/2407.19875

>MaskInversion: Localized Embeddings via Optimization of Explainability Maps
https://walidbousselham.com/MaskInversion/

>Task-Adapter: Task-specific Adaptation of Image Models for Few-shot Action Recognition
https://arxiv.org/abs/2408.00249
>>
File: tmp_94tfa0d.png (662 KB, 768x1024)
662 KB
662 KB PNG
>>
https://civitai.com/models/599651/forcutexl
https://civitai.com/models/490622
Where can I get more of these 12GB SDXL models? I want to compare if they're better than the 6GB SDXL models.
>>
File: FD_00058_.png (1.16 MB, 1024x1024)
1.16 MB
1.16 MB PNG
>>
File: Flux_1722737982_0001.png (1.21 MB, 896x1152)
1.21 MB
1.21 MB PNG
>>
File: FLUX__00015_.png (1.03 MB, 1024x1024)
1.03 MB
1.03 MB PNG
has it been 2 days already
I should probably get some sleep
>>
File: tmp_wg4xzj6.png (766 KB, 768x1024)
766 KB
766 KB PNG
>>
File: scribble_challenge010.jpg (501 KB, 2112x1062)
501 KB
501 KB JPG
challenge continues. my drawing sucks.
>>
File: 1694480287922908.png (1.07 MB, 1024x1024)
1.07 MB
1.07 MB PNG
>Trump with a cutout of Hatsune Miku
>>
File: FLUX__00047_.png (1.16 MB, 896x1152)
1.16 MB
1.16 MB PNG
woah meta
>>
File: FD_00064_.png (1.27 MB, 1024x1024)
1.27 MB
1.27 MB PNG
>>101713360
we did the same thing
>>
File: 1696851126764633.png (1.32 MB, 1024x1024)
1.32 MB
1.32 MB PNG
this one turned out pretty good:
>>
File: 1714108670447802.png (984 KB, 1554x768)
984 KB
984 KB PNG
>>
File: ComfyUI_01336_.png (952 KB, 1200x768)
952 KB
952 KB PNG
Flux is fun
>>
am I alone in feeling like flux is overhyped right now and not that great? I know we've had a drought so I get why everyone's excited to play with the new toy, but I hardly see this being the future of local..
>>
>>101713498
compared to what
>>
>>101713498
It's an open sourced model with great text and prompt adherence. That's basically what everyone wanted since SD 1.5
>>
>>101713498
It's better than Stable Diffusion run of the mill. It offers a much out of the box experiment in my opinion. Although I said it can't be fine tuned, not sure if it's true, which would be a big downside.
>>
>>101713508
why's it got to be compared to something? just because we don't have anything superior at prompt comprehension doesn't make it an optimal outlet for local's future.
>>
File: BMP_FLUX_00002_.png (1.55 MB, 1024x1024)
1.55 MB
1.55 MB PNG
>>
>>101713498
The future is bigma.
>>
>>101713526
One thing I've learned through the years is, never underestimate autists and furfags. Pretty sure someone will discover a way to finetune it.
>>
File: 00339-187241492.png (493 KB, 768x432)
493 KB
493 KB PNG
>>101713498
>but I hardly see this being the future of local
compared to what?
The previous SD models were all shit before finetunes. Flux is actually not, I could easily mistake sd general for the dall-e3 general now if i didn't know any better.
Yeah, promptlets here got hyped at each release and flooded the threads with lower quality images than usual until finetunes started coming out and anons calmed down. This time is definitely very different.
>>
>>101713528
Because I want to know where you think the bar is, and why flux is "overhyped".
>>
File: 1696038227096350.png (1.57 MB, 821x821)
1.57 MB
1.57 MB PNG
>>101713528
>we don't have anything superior at prompt comprehension
ALL models, local or not, suck at prompt comprehension. I'm tire of pretend we have any model that could at understand more than 2 or 3 sentences without fucking it up. I'm pretty sure that in order for us to achieve better prompt comprehension, they will have to change something on how this technology work.

Maybe the model starting to have some conceptual idea of what each thing is on the scene and building a 3D model of the thing.
>>
>>101713512
sure but it seems poorly packed for the amount of params and severely lacking on a stylistic front, lots of concepts it doesn't know or understand. I realize it's a base model, but considering the compute and difficulty that has to go into fine-tuning it (if we even can fine-tune the distilled models), and the amount of params it's using, it feels really lackluster. like they half assed it and we'd have been just as impressed if anyone else had trained and released at the same param amount
>>
File: BMP_FLUX_0003_.png (1.57 MB, 1024x1024)
1.57 MB
1.57 MB PNG
>>
File: 1721835321683311.png (559 KB, 445x597)
559 KB
559 KB PNG
>>101713571
I want an Indian woman with a purple Celtic sword that seems to be made with glass fighting a blue dragon, and she's riding a green Pikachu. The Pikachu has wings (the wings, and only the wings, are red), and the woman is wearing a Chilean soccer team t-shirt from the 1970s. Also, there's an old Japanese man with a moustache dressed as the "Joker" in the background, sitting on a blue bench and reading a newspaper. The man is talking to a Black woman dressed like Batman. Oh, and the whole scene takes place on Mars. And there's a flying McDonald's in the sky that you can see further in the background, and there's a giant sign that says 'Free iPads!' on it
>>
File: namekian laugh (1).gif (1.41 MB, 320x240)
1.41 MB
1.41 MB GIF
>>101713528
>just because we don't have anything superior at prompt comprehension doesn't make it an optimal outlet for local's future.
YES IT DOES
At least, until something even better comes along.
>>
File: ComfyUI_10275_.png (1.66 MB, 1120x1440)
1.66 MB
1.66 MB PNG
>>101713498
new ai toys are always popular until people get bored of its limitations, for example, img2video, i've paid for luma and gen-3 alpha and they are both shitte, no wonder Sora was trashed too, its literally like playing slots in the casino and see how your credits dwell down, yeah flux is great but i think most people gens looks all the same, the real holy grail of ai development is consistency
>>
File: tmp50kdg_i5.png (929 KB, 768x1024)
929 KB
929 KB PNG
>>
File: 00073-1983335564.png (3.38 MB, 1920x1280)
3.38 MB
3.38 MB PNG
While I'm sure you people are busy experimenting with boobs, how about backgrounds?

Please try prompting:

- an endless prairie
- dark forest with dirt path
- labyrinthine ornate stone ruins and a flowing creek
- mountainside brick road
- sci-fi cityscape
- log cabin, cliffside, rock wall, diverse flora, solo tree

And show us all what Flux can do, pretty please, thanks.
>>
File: BMP_FLUX_00016_.png (1.72 MB, 1024x1024)
1.72 MB
1.72 MB PNG
>>101713670
I'm just trying some old prompts and man does it love to make out of focus backgrounds
>>
>>101713692
Just neg-- oh, wait...
>>
My 3060 12 gb is getting absolutely buck broken with flux. Any idea when optimizations are coming?
>>
File: delux_con_00018_.png (1.5 MB, 1216x832)
1.5 MB
1.5 MB PNG
>>101713309
leg arm has the most soul

>>101713485
oh shit, 50% off was today??

>>101713498
its like when you've had a crush on a girl for a whole year (sd3) and you finally score a date with her and you find out she's pretty retarded and smells weird. bummed and dejected, some flashy bimbo gives you attention randomly (flux) and you have a fun weekend together.
is it deep? will it last? who knows, but its fun when you thought everything sucked

>>101713535
welcome to the flux club

>>101713715
ait is potentially coming. or not. we'll never know. its also possible the model can be quantized but don't hold your breath
>>
>>101713715
same gpu, but my cpu is taking a pounding because it's picking up the slack
2.5 minutes a pop is tolerable, but if the anatomy was fucked on each gen then it would wear thin VERY fast
>>
File: 1699258484083155.png (1.14 MB, 1024x1024)
1.14 MB
1.14 MB PNG
Miku 2024
>>
>>101713692
that's why I ask for backgrounds only, NO PEOPLE, obviously, or you can't assess its true capabilities with just scenery
>>
File: BMP_FLUX_00084_.png (1.55 MB, 1024x1024)
1.55 MB
1.55 MB PNG
>>101713761
Sorry anon. I'm busy experimenting with boobs.
>>101713715
Really? 3070 8gb Vram and 32 Ram and it seems fine. Did you download the low Vram version?

My biggest complaint so far is it likes to blend clothing into skin textures in weird ways for like 20-30% of gens so far but I'm just using the basic bitch workflow.
>>
File: FD_00114_.png (1.52 MB, 1024x1024)
1.52 MB
1.52 MB PNG
>>101713670
>- an endless prairie
>>
File: FD_00115_.png (1.33 MB, 1024x1024)
1.33 MB
1.33 MB PNG
>>101713670
>- dark forest with dirt path
>>
File: FD_00116_.png (2.09 MB, 1024x1024)
2.09 MB
2.09 MB PNG
>>101713670
>- labyrinthine ornate stone ruins and a flowing creek
>>
File: FD_00117_.png (1.9 MB, 1024x1024)
1.9 MB
1.9 MB PNG
>>101713670
>- mountainside brick road
>>
File: ffhffghg.jpg (1.52 MB, 1293x3071)
1.52 MB
1.52 MB JPG
>>
>>101713670
>- sci-fi cityscape
>>
File: FD_00118_.png (1.58 MB, 1024x1024)
1.58 MB
1.58 MB PNG
>>101713949
forgot to attach pic
>>
File: ffhfghgff.jpg (1.38 MB, 1288x3072)
1.38 MB
1.38 MB JPG
no graveyard...
no classroom...
no cactus garden...
no wet slimy gooey cave interior...
well they weren't requested.......
>>
File: FD_00119_.png (2.2 MB, 1024x1024)
2.2 MB
2.2 MB PNG
>>101713670
>- log cabin, cliffside, rock wall, diverse flora, solo tree
>>
>>101713976
>>101713934
What the fuck is wrong with your pics, why are they all washed out?
>>
File: BMP_FLUX_00102_.png (1.57 MB, 1024x1024)
1.57 MB
1.57 MB PNG
>>101713990
Living the dream
>>101714002
I actually like them
>>
File: jgjgjhh6u.jpg (2.24 MB, 2048x3072)
2.24 MB
2.24 MB JPG
>>101714002
muh style, but i feared they were a bit too...ill adjust
>>
File: tmpxo4z19s1.png (897 KB, 768x1024)
897 KB
897 KB PNG
>>
File: cosplay.png (1002 KB, 1024x1024)
1002 KB
1002 KB PNG
>>101713360
>>101713415
>>101713460
>>101713746
>>
File: BMP_Bama.png (1.47 MB, 896x1152)
1.47 MB
1.47 MB PNG
>>101714039
>>
File: tmpt907vmfu.png (1.07 MB, 768x1024)
1.07 MB
1.07 MB PNG
>>
File: 1716981186329040.png (1.13 MB, 1024x1024)
1.13 MB
1.13 MB PNG
>>
File: BMP_FLUX_00130_.png (1.42 MB, 1024x1024)
1.42 MB
1.42 MB PNG
AW HELL NAW, wtf is this artstyle
>Potato Christ in the background
>>
File: tmprtc4z5t1.png (1.02 MB, 768x1024)
1.02 MB
1.02 MB PNG
>>
>>101714152
>Flux is anti avatar faggot
can this model get any better?
>>
File: ComfyUI_Flux_231.png (1.38 MB, 1024x1024)
1.38 MB
1.38 MB PNG
>>
Flux is 80% close to Dalle3. Dalle3 still mogs it but it's almost there and you can do it uncensored so it wins for me
>>
blackpilled again. SD1.5 still undefeated for 3DPD tits
>>
File: FD_00133_.png (1.04 MB, 1024x1024)
1.04 MB
1.04 MB PNG
>>101714180
Only if you don't know how to prompt without LoRAs. I can consistently get this woman.
>>
File: 1722743617.png (1.16 MB, 1024x1024)
1.16 MB
1.16 MB PNG
>>
>>101713233
12GB models are used for finetuning. It won't matter for regular genning.
>>
File: 00017_.png (1.7 MB, 1024x1024)
1.7 MB
1.7 MB PNG
>>101713990
Reminds me a bit of one I made earlier
>>101710102
only it was 'small cabin on the edge of a planet.'
This is 'red barn at the end of the universe.'
>>
File: delux_con_00017_.png (921 KB, 1216x832)
921 KB
921 KB PNG
first time using comfyui on a macbook and I'm rather upset that it doesn't handle touch events correctly

>>101714180
now I wanna try to make debos

>>101714190
less accurate but less dogged
>>
File: BMP_01706_.jpg (914 KB, 1536x1536)
914 KB
914 KB JPG
>>101714180
I'm pretty sure it's just anti-me. I have plenty of shit gens from other models I've never shared
>>
File: delux_flebo_00001_.png (1.32 MB, 1216x832)
1.32 MB
1.32 MB PNG
>>101714180
it debo'ed first try. flux rox
>>
File: tmpjmnrzyt7.png (1.01 MB, 768x1024)
1.01 MB
1.01 MB PNG
>>
>>101713498
It's fucking useless if people can't make character LORAs for it.
>>
>>101714188
>no helicopter in the upper left corner
trash/10
>>
>>101714284
It's basically dead to me if it can't draw cute anime girls from the next season of anime.
>>
File: Flux_00304_.png (1011 KB, 1024x768)
1011 KB
1011 KB PNG
>>
File: delux_flebo_00003_.png (1.41 MB, 1216x832)
1.41 MB
1.41 MB PNG
I'm flebo'ing my gourde off rn
>>
>>101714338
did you just prompt "___ in the style of world of warcraft"? looks neat
>>
>>101714180
the opposite, it's made for avatars, very similar to dalle in that sense.
>>
File: delux_flebo_00002_.png (1.14 MB, 1216x832)
1.14 MB
1.14 MB PNG
>>101714338
its actually mindblowing how near-perfect the UI is
>>
>>101714336
LoRA training is out and working. Just need 40gb of vram.

You're not poor... are you?
>>
>>101714359
you mean the guy that said he was training? who didn't produce any results?
>>
>>101714359
It's basically dead to me if LORAs of next season's anime girls aren't posted on Civitai within the first few weeks of the season.
>>
File: 1702423342598153.png (1.16 MB, 1024x1024)
1.16 MB
1.16 MB PNG
>>
>>101714371
yes you believe the guy who hasn't even tried who says you can't?
>>
File: ComfyUI_170325_.png (1.07 MB, 1024x1024)
1.07 MB
1.07 MB PNG
https://comfyanonymous.github.io/ComfyUI_examples/flux/#simple-to-use-fp8-checkpoint-version

If anyone needs it here's a single file checkpoint version of flux-dev that you can load with the regular "Load Checkpoint" node.
>>
>Mouse can't gen his waifu while Debo can spam his shota persona with ease
How will BMP ever recover after this?
>>
File: Flux_00306_.png (1021 KB, 1024x768)
1021 KB
1021 KB PNG
>>101714351
>>101714356

>World of Warcraft mage ingame screenshot, UI and HUD visible,

Capitalizing World of Warcraft helps.
>>
>>101714398
Just fp8 though
>>
>>101714383
I don't believe anyone until they provide me evidence. a screenshot of vram usage isn't convincing me of shit
>>
>>101714415
>literally looks like icecrown
>xp bar
neat
>>
File: delux_flebo_00007_.png (1.25 MB, 1216x832)
1.25 MB
1.25 MB PNG
>>101714414
some people refuse to acknowledge I'm one of the best prompters. thats why I'm so eager to help people. whats the point in having all this knowledge if I can't share it, yknow?
>>
File: Flux_00308_.png (1.09 MB, 1200x1024)
1.09 MB
1.09 MB PNG
>>
File: delux_flebo_00009_.png (1.3 MB, 1216x832)
1.3 MB
1.3 MB PNG
>>101714415
>Capitalizing World of Warcraft helps.
thats fascinating, actually. would be interesting to do some XY grids comparing capitalization of various things
>>
File: Flux_00309_.png (1.18 MB, 1200x1024)
1.18 MB
1.18 MB PNG
rate my transmog
>>
File: delux_flebo_00010_.png (1.3 MB, 1216x832)
1.3 MB
1.3 MB PNG
>>101714476
can we get a necromancer obama raising a biden minion to fight for him?
>>
File: Flux_00313_.png (953 KB, 1024x768)
953 KB
953 KB PNG
>>101714497
Forgot to prompt Joe Biden but lol
>>
>>101714422
Everything else so far has proven to be trainable yet you believe a screenshot from a discord tranny because it's in screenshot form? Is that really how your brain works?
>>
File: delux_flebo_00011_.png (1.25 MB, 1216x832)
1.25 MB
1.25 MB PNG
>>101714518
>forgot to prompt joe biden
>still gets an army of joe bidens
another flux W
>>
>>101714519
NTA, but I think a certain level of not adhering to regular rules is an acceptable thing to assume in this case. Nothing has been quite this good or this large before. Sure, it's likely it can be magicked into training, but we also can't just assume that it is possible without breaking the model.
>>
File: 00268-553801033.jpg (800 KB, 1560x2328)
800 KB
800 KB JPG
>>
File: F_00003_.jpg (313 KB, 1024x1024)
313 KB
313 KB JPG
>>
File: BMP_FLUX_00141_.png (1.61 MB, 1024x1024)
1.61 MB
1.61 MB PNG
>>101714414
I hope you liked whatever he genned anon, because I'm sure he'll be posting it 40 times a day for the next three weeks or so
>>
>>101714518
the real lich king
>>
File: Flux_00317_.png (890 KB, 1024x768)
890 KB
890 KB PNG
>>
File: Flux_00319_.png (843 KB, 1024x768)
843 KB
843 KB PNG
If someone can get it to actually produce joe biden minions, I applaud you.

>World of Warcraft Barack Obama Necromancer avatar ingame screenshot, J(oe Biden:1.1) minions following Barack Obama, UI and HUD visible, nameplates on top right reads "Obongo, third person top down perspective
>>
>>101714598
>J(oe Biden:1.1)
?
>>
>>101714604
oops
>>
File: ComfyUI_170338_.png (1023 KB, 1024x1024)
1023 KB
1023 KB PNG
>>101714420
It's for people who have normal hardware. If you have good hardware you can handle the multiple files.
>>
>>101714545
>>101714545
And I think a certain level of shut the fuck up needs to some into play when retards come in here spouting shit they don't understand because they saw a screenshot that said no.
It's absolutely insane to me that they do this. It's got to be SAI shills trying to muddy the waters.
>>
File: tmpsgo8asik.png (875 KB, 768x1024)
875 KB
875 KB PNG
>>
File: 1720788499689604.png (1.06 MB, 1024x1024)
1.06 MB
1.06 MB PNG
>>
File: Flux_00321_.png (932 KB, 1024x768)
932 KB
932 KB PNG
>>
File: FrGBNCq-YmzwkTFO8QdyC.png (1011 KB, 768x1024)
1011 KB
1011 KB PNG
I got sidetracked by this whole flux thing and didn't finish my 10 thread recap. And to make matters worse I accidentally closed my browser and lost all my progress (I was half done)

maybe I start over, to punish myself....
>>
File: delux_flebo_00013_.png (1.17 MB, 1216x832)
1.17 MB
1.17 MB PNG
>>101714563
we could be friends but you playin

>>101714617
do you have any insights into the whole "is flux trainable" debate happening?
>>
File: BMP_FLUX_00249_.png (957 KB, 1024x1024)
957 KB
957 KB PNG
>>
FLUX dev is better at making Quokkas with a bit more instructions in the prompt but the ears still come out a bit too big sometimes
>>
File: 1706488070905609.png (927 KB, 1024x1024)
927 KB
927 KB PNG
Diablo 2 ingame screenshot, UI and HUD visible, the main character is Miku Hatsune

it's not 2 but it still looks neat
>>
File: 1710422511132252.png (1.32 MB, 1024x1024)
1.32 MB
1.32 MB PNG
>>101714667
i'd literally play this over 4
>>
File: 1703240823114630.png (1.18 MB, 1024x1024)
1.18 MB
1.18 MB PNG
no fucking way, it knows what KCD is and that it's a medieval game

>Kingdom Come Deliverance ingame screenshot, UI and HUD visible, the main character is Miku Hatsune
>>
File: BMP_FLUX_00251_.png (1012 KB, 1024x1024)
1012 KB
1012 KB PNG
>>101714678
Same
>>
File: 00024.png (784 KB, 1024x1024)
784 KB
784 KB PNG
>>
File: delux_obobo_00002_.png (894 KB, 1216x832)
894 KB
894 KB PNG
>>101714598
seeing what I can get. so far, I got this weird sims hybrid
>>
File: Flux_00327_.png (847 KB, 1024x768)
847 KB
847 KB PNG
>>101714678
>>101714667
>>
>>101714686
it knows alot, llms are really good at tagging screenshots and such, the issue with the model is mostly related to artists, styles and weeb stuff.
>>
File: ComfyUI_170228_.png (897 KB, 1024x1024)
897 KB
897 KB PNG
>>101714639
Dev is definitively trainable it just might take some creativity to get it training on a 4090.
>>
>>101714713
If that tagger doesn't know who megumin is even though she's in the dataset, that's a problem.
>>
File: delux_obobo_00003_.png (906 KB, 1216x832)
906 KB
906 KB PNG
what the fuck
>>
>>101714724
worst pet class yet?
>>
>>101714705
>>101714667
What's are words to describe what diablo 2 looks like? Isometric computer game? Isometric gothic horror?
>>
>>101714741
could prompt isometric view, for >>101714667
I didnt specify a view
>>
>>101714741
I used orthographic view
>>
File: BMP_FLUX_00267_.png (928 KB, 1024x1024)
928 KB
928 KB PNG
Cursed
>>
File: Flux_00332_.png (1.25 MB, 1024x768)
1.25 MB
1.25 MB PNG
>>
The think I like about flux is that I often do have to double check if it's AI or not because the images themselves often aren't very AI slopped.
>>
Is there a good yolov8 (ADetailer) model for anime person segmentation? Ultralytics has a person segmentation model but it only works well on realistic images and absolutely shits itself on anime.
>>
File: delux_obobo_00006_.png (823 KB, 1216x832)
823 KB
823 KB PNG
he got a cute lil necromancer tie and a cute lil press secretary minion. he's not powerful enough to summon biden
>>
>>101714774
Now we're getting somewhere.
Might get more detail with a room focus like "sacrificial ritual chamber" or something similar.
>>
Guys I had an idea
https://files.catbox.moe/bdr8uh.png
>>
>>101714801
why is that interesting?
>>
File: delux_obobo_00007_.png (937 KB, 1216x832)
937 KB
937 KB PNG
Channeling: Drone Strike
>>
File: Flux_00342_.png (957 KB, 1024x768)
957 KB
957 KB PNG
I moved on to GTA5 miku
>>
>>101714858
Do Postal Miku
>>
File: 1698002977570098.png (1.15 MB, 1024x1024)
1.15 MB
1.15 MB PNG
>>
File: delux_obobo_00008_.png (821 KB, 1216x832)
821 KB
821 KB PNG
>>101714858
there's an upcoming gta gacha yokai mobile anime waifu collector that I'm weirdly interested in. you can be a catgirl and drive a porsche
>>
File: 1717815138361613.png (1.24 MB, 1024x1024)
1.24 MB
1.24 MB PNG
said warcraft 3 but I got WoW
>>
>>101714885
Might have to specify more since WoW probably has way more data. Warcraft the Frozen Throne might work better.
>>
File: 1710472380976409.png (1.32 MB, 1024x1024)
1.32 MB
1.32 MB PNG
>>101714885
RTS game warcraft 3 is closer:
>>
File: ComfyUI_01481_.png (1.17 MB, 1024x1024)
1.17 MB
1.17 MB PNG
>>101714335
It doesn't know what to do with the logo though
>>
File: file.jpg (254 KB, 1792x1024)
254 KB
254 KB JPG
>>101713727
>ait is potentially coming. or not. we'll never know
im working on it but id rather not be desu
soon:tm: trust the plan etc
>>
File: ComfyUI_00357_.png (1.53 MB, 1280x720)
1.53 MB
1.53 MB PNG
>>101713692
My solution to this was to describe everything in the background and slap the word "detailed" everywhere I can. And if that's not enough then I prompt for various art styles that don't influence the image too much. Funnily enough image related was putting "by Pablo Picasso" in the prompt.
>>
File: ComfyUI_00356_.png (1.16 MB, 1280x720)
1.16 MB
1.16 MB PNG
>>101714916
>>101713692
This is what I get without the style specified.
>>
File: delux_obobo_00010_.png (982 KB, 1216x832)
982 KB
982 KB PNG
>when your biden summoning ritual misfires and you summon mike pence instead

>>101714904
I trust

>>101714916
can you make this gen with a dystopian cyberpunk city? its got bladerunner vibes without being in bladerunner
>>
File: Flux_1722748228_0001.jpg (2.01 MB, 1792x2304)
2.01 MB
2.01 MB JPG
>Flux doesn't know who Max Verstappen is
>sad.png
>>
File: ComfyUI_00363_.png (1.6 MB, 1280x720)
1.6 MB
1.6 MB PNG
I also like how "Drawn in pencil. Colored." looks. If you switch that out for "Sketched", it's even rougher which can give a more interesting vibe.
>>
>>101714931
it's hilarious and amazing that we can actually generate WoW characters in the style of the game, I wonder how many games it can actually understand
>>
>>101714969
Someone needs to do Skyrim
>>
File: 1711730038713601.png (1.22 MB, 1024x1024)
1.22 MB
1.22 MB PNG
a wild Miku appears:
>>
>>101714969
>WoW characters in the style of the game
That's not in the style of the game though, it just looks like Sims or 2nd life characters
>>
File: ComfyUI_01486_.png (1.57 MB, 1024x1024)
1.57 MB
1.57 MB PNG
>>101714902
Prompted it on the left now
>>
File: 1703772731328504.png (1.29 MB, 1024x1024)
1.29 MB
1.29 MB PNG
put your faith in the juice
>>
File: can_be_trained.png (32 KB, 825x159)
32 KB
32 KB PNG
So apparently scripts to train Flux are ready. The author claims that DoRA training works well. Unsure about full model finetuning for obvious reasons.
>>
gonna try to get in on this with SD3 to see how they compare
>>
File: delux_obobo_00017_.png (1.13 MB, 1216x832)
1.13 MB
1.13 MB PNG
>>101714969
it can be hit or miss. I was trying both castlevania and pokemon yesterday and struggling. maybe I'm a promptlet though, idk

>>101714971
I'll do you one better, I'm cooking up some morrowind obamas next. I actually did morrowind obama all the way back in SD1.5 2022
>>
File: FLUX__00058_.png (1.38 MB, 1024x1024)
1.38 MB
1.38 MB PNG
No fuckin way
A muscular man with big, black spiky hair. He is wearing an orange gi with a blue belt. He is standing next to a wooden hut on a small island with ocean all around.

I wanted a realistic dude but it thought it knew better
>>
File: 1722748065.png (951 KB, 1024x1024)
951 KB
951 KB PNG
>>
>>101715048
prompt?
>>
>>101715051
beautiful blonde woman wearing translucent organza pink dress and black high heels and frosted pink lipstick and blue eyeshadow and bra and panties and pink nail polish standing on a rooftop
>>
File: file.jpg (331 KB, 1024x1792)
331 KB
331 KB JPG
>>101714931
atm it should be good for 24gb cards without being split, ill have to double check the workspace for the entire model with fp8 weights
for split and in general the next steps are
- workspace lazy loading/free after use, this will allow more (or maybe all) modules with weights to be loaded at once, reducing the time between layers, or for full model it will allow other (pytorch) models/other modules to be loaded without so much juggling
- module(s) for the first and last part of the model
- wrt above those parts could just be pytorch and integrate, but thats more complex and depends whether it can be done as a comfy node (probably not)
- put it all together
- verify numerical accuracy
- test speed of set_constant method
- op performance tuning for any potential extra speed
- rewrite ait comfy node (and keep rewriting it whenever comfy breaks it)
maybe more, idk, im tired ive barely slept
>>
File: BMP_FLUX_00312_.png (1.66 MB, 1024x1024)
1.66 MB
1.66 MB PNG
>>101714971
>>
>>101715090
DOVAHKIIN DOVAHKIIN NAAL OK ZIN LOS VAAH RIN
>Miku's voice
>>
File: delux_mobama_00001_.png (1.15 MB, 1216x832)
1.15 MB
1.15 MB PNG
I can't get it to give me low-poly

>>101715083
>ill have to double check the workspace for the entire model with fp8 weights
there's prob a lot more people on fp8 weights now. hopefully there's some good mileage gain in there
>im tired ive barely slept
I thought this was a weird hour for you but wasn't sure
>>
>>101714686
It knows that a game called Kingdom Come Deliverance is a medieval game? Woa!
>>
sd3 seems to struggle to distinguish the player character from other NPCs in view
>>
File: file.jpg (449 KB, 1024x1792)
449 KB
449 KB JPG
>>101715089
thanks for reminding me of another step
- accept that ill receive zero recognition, only hateful comments
>>101715103
its 6am and i went to sleep about 2
i'd rather be putting together a dataset of low-poly stuff and everything else thats clearly missing desu
>>
>>101715123
I mean, it's a RPG but the aesthetics of the generation are surprisingly close to the actual game, which is in the 14th century, so the dataset must have images of it
>>
>>
File: FLUX__00062_.png (1.09 MB, 896x1152)
1.09 MB
1.09 MB PNG
>>
File: delux_mobama_00008_.png (912 KB, 1216x832)
912 KB
912 KB PNG
>>101715110
using dev. the prompt is using wildcards
Elder Scrolls Morrowind character Barack Obama wearing {Acrobat | Agent | Archer | Assassin | Barbarian | Bard | Battlemage | Crusader | Healer | Knight | Mage | Monk | Nightblade | Pilgrim | Priest | Rogue | Scout | Sorcerer | Spellsword | Thief | Warrior | Witchhunter} gear, ingame screenshot from Morrowind, {Altmer | Argonian | Bosmer | Breton | Dunmer | Imperial | Khajiit | Nord | Orc | Redguard} Obama and Joe Biden adventure at {Ald'ruhn | Balmora | Caldera | Dagon Fel | Ebonheart | Gnisis | Hla Oad | Khuul | Maar Gan | Molag Mar | Pelagiad | Sadrith Mora | Seyda Neen | Suran | Tel Aruhn | Tel Branora | Tel Mora | Tel Vos | Vivec | Vos}, UI and HUD visible, third person adventure perspective Morrowind The Elder Scrolls, __angle__ low poly playstation 1 graphics akin to early 2000s, blocky and jagged polygons

the flat prompt is
Elder Scrolls Morrowind character Barack Obama wearing  Rogue  gear, ingame screenshot from Morrowind,  Argonian  Obama and Joe Biden adventure at  Ebonheart , UI and HUD visible, third person adventure perspective Morrowind The Elder Scrolls, atmospheric perspective low poly playstation 1 graphics akin to early 2000s, blocky and jagged polygons


>>101715132
this reminds me strongly of lineage 2 for some reason

>>101715135
>accept that ill receive zero recognition, only hateful comments
hate only has power when you grant it. thats just the same guy on repeat because he realized he can get under your skin. tiny people like that are easy to ignore
>>
>>101715155
Maybe they captured an image of the entire webpage and used that as grounding for the LLM to then caption each individual image on the page.
>>
File: 1698882115086587.png (1.05 MB, 1024x1024)
1.05 MB
1.05 MB PNG
Miku leads all DPS charts
>>
>>101715137
The aesthetic looks nothing like KC:D
>>
>>101713099
Because I am retarded, I will ask:

What even are the correct names for these?

Like what do I call the app that runs the model file?

And what even do we call in general models like sd and dalle - vs the text ones?

And can I run a visual model in llama.cpp? Is that retarded to ask?
>>
>>101715173
>DPS
What's that mean?
>>
>>101715193
damage per second
>>
>>101715189
>text 2 image models? image generation models?
I believe 'Diffusion Models' would be more accurate
>>
File: 00031.png (1.34 MB, 1024x1024)
1.34 MB
1.34 MB PNG
>>
>>101715207
why not?
>>
>>101715193
Diddling Pedo Scum
>>
File: 00032.png (1.35 MB, 1024x1024)
1.35 MB
1.35 MB PNG
>>
File: delux_mobama_00009_.png (1.11 MB, 1216x832)
1.11 MB
1.11 MB PNG
>>101715181
>what do I call the app that runs the model file?
'apps' are fine. 'frontend' is somewhat common too
>what even do we call in general models like sd and dalle
imggen models. can also refer to them as 'text-to-image'
>>
>>
>>101715201
aha, I recommend spraypaint for that.

>>101715218
Why are people so down on average child rapists? They don't even perform circumcision on them, typically.
>>
as it turns out, teh new model is fodder for the rich and fanciful, for the idle rich. down with Flux, and down with the petite bourgeoisie. you can shove your 4090 up your fancy rich ass. aka eat shit and die.
>>
>>101715225
Explain all of those lingo things, I have no idea what that is, except I know that Flux makes big tits, so I'm in.
>>
File: delux_mobama_00010_.png (933 KB, 1216x832)
933 KB
933 KB PNG
>>101715251
it can run on 8gb tho
>>
>>101715262
shut the fuck up nigger. the new model is only for the rich. countersignaling this message only proves you're one of THEM
>>
File: F_00026_.png (1.08 MB, 1024x1024)
1.08 MB
1.08 MB PNG
>>101715173
I tried elder scrolls online, it seems to know the "style"
>>
>>101715251
isn't it always
>>
>>101715225
It still fits in the broad category of diffusion models tho
>>
i hate flux because i am poor. many such cases
>>
>>101715269
>be me
>not rich
>running flux
cease your whining faggot
>>
File: 1704285179976727.png (1.25 MB, 1024x1024)
1.25 MB
1.25 MB PNG
pretty sure it thinks the game is in first person so I said miku standing in front of the player for a skyrim ingame prompt
>>
i am poor because my brain has a lot of defects. and so i hate those who's brains are correctly ordered. i will burn the world to cinders just beacause
>>
>>
File: 1722750640.png (923 KB, 1024x1024)
923 KB
923 KB PNG
>>
>>101715301
Is that Donald Biden?
>>
File: Untitled.png (1 KB, 260x29)
1 KB
1 KB PNG
>>101715262
>can
Maybe I'm doing it wrong but I can't imagine anyone waiting this long per image. Never used comfy before.
>>
>>101715314
yeah you are poor. just have a trust fund bro, just get a high end card bro. what the fuck is wrong with you?
>>
File: FLUX__00192_.png (1.22 MB, 896x1152)
1.22 MB
1.22 MB PNG
>>
File: delux_mobama_00011_.png (1.04 MB, 1216x832)
1.04 MB
1.04 MB PNG
>>101715269
its not just for the rich though. the middle class can also use it with some struggle. I can't dock BFL points because their model still runs on a lot of consumer hardware. you'd have a better argument if you were attacking the pro model which we won't get and probably can't run. you can also make a good argument in attacking nvidia for capping consumer vram capacity even though they've known AI was coming for years.
>>
>>101715329
no, not all of us can afford a 4090. 1.5 was good because it was runnable by just about everyone. all AI dev leads inevitability towards only major corps with boundless infrastructure. you are dust
>>
>>101715251
>as it turns out, teh new model is fodder[snip]
I'll put this here into kunoichi-dpo-v2-7b.Q6_K.gguf and see what it says (1000 character limit):
as it turns out, teh new model is fodder for the rich and fanciful, for the idle rich. down with Flux, and down with the petite bourgeoisie. you can shove your 4090 up your fancy rich ass. aka eat shit and die.

in the meantime, i've got my 3090ti, and that's more than enough for me.

yeah, the 4090 is a beast. no doubt about it. but it's also priced like a beast. for the rest of us, there's always the 3090ti. or even the 3080ti, for those on a budget.

and for those who really want the latest and greatest, well, they can wait a year or two for the next generation of cards. by then, the prices will have come down, and the performance will have increased. it's the circle of life.

so, in conclusion, the 4090 is not for everyone. and that's okay. there are plenty of other options out there that can satisfy your gaming needs. just because something is the latest and greatest doesn't mean you need to have it.

i, for one, am happy with my 3090ti. it's more than capable of handling anything i throw at it, and i don't feel the need to upgrade just because there's a new kid on the block.

as for Flux, well, i'm sure they'll come out with something new and exciting eventually. but for now, i'm content with what i have. and that's the most important thing.

so, in a nutshell, the 4090 may be the cream of the crop, but it's not the only option out there. there are other cards that can provide excellent performance without breaking the bank. and at the end of the day, it's all about what works best for you.

so, let's not get too caught up in the hype. let the rich and fanciful have their 4090s. for the rest of us, there's plenty of other great options out there. [end of text]
>>
File: file.png (103 KB, 1124x368)
103 KB
103 KB PNG
nice try faggot
>>
File: Flux_00001_.png (1.63 MB, 1024x1024)
1.63 MB
1.63 MB PNG
>>101715321
Yeah I tried out dev first.
Took 614 seconds.
Trying Schnell next.
>>
>>101715355
I need to watch the page master again
>>
File: FLUX__00068_.png (959 KB, 896x1152)
959 KB
959 KB PNG
>>
File: file.png (378 KB, 1421x456)
378 KB
378 KB PNG
>>101715341
hurr durr i am gay.
2k for a 3090. eat shit and die nigger
>>
File: FD_00172_.png (1.53 MB, 1024x1024)
1.53 MB
1.53 MB PNG
>>101715031
didn't work for krillin
>>
File: 1696760060437216.png (1.22 MB, 1024x1024)
1.22 MB
1.22 MB PNG
an ingame screenshot of Skyrim, UI and HUD visible. Miku Hatsune is standing in front of the player and she is holding a large greatsword. Miku is wearing shiny, reflective plate armor. She is facing the camera. She is near a local medieval town with shops. The background is Whiterun from Skyrim.

could prob take the hud stuff out but neat
>>
>>101715314
Can you explain a bit how you are doing it? I want to try. What card do you have? I have a 6950xt.
>>
>>101715368
Just pay it off over 2 years bro
>>
File: delux_mobama_00017_.png (1.07 MB, 1216x832)
1.07 MB
1.07 MB PNG
>>101715338
>not all of us can afford a 4090
I'm doing ok on a 4070. not the best, not the worst, but I can cook.
>1.5 was good
I agree. I'll go even further and say 1.5 is legendary. it will always have some relevance just because it can be run by almost anyone. but that comes at a cost of capped and aging performance. the nature of progress is that you'll always need more
>all AI dev leads inevitability towards only major corps
I hope not. that's why we're here in sdg instead of de3. we want to believe in local, open-source models building the future. thats why I'm resonant with the argument that flux is bad if it isn't trainable or that open-source isn't a committed mission of BFL. I think you're warranted in your doomerposting but I also think you're premature
>>
File: file.png (404 KB, 1397x400)
404 KB
404 KB PNG
>>101715380
could buy this piece of reconstructed shit for that much
you deserve to die
>>
>>101715368
>>101715338
Usually you build an unoptimized blob. Then you proceed to optimize it. Stop being a salty whiny faggot and be patient.
>>
>>101715393
you have never been poor. you are a temporarily embarrassed millionaire. eat shit and die
>>
meds
>>
>>101715380
>2 years
I can buy 3 of these per paycheck. I also can't afford a house. you'll call me rich and you'd be wrong
>>
>>101715376
>amd
Are you running linux? According to https://github.com/comfyanonymous/ComfyUI it looks like comfy only works with AMD on linux.
I have nvidia so I have no clue where you should start.
>>
>>101715372
Half your prompt doesn't apply but okay
>>
>>101715393
>bro they'll optimize it bro
AI bros who know nothing about AI are so annoying
>>
>>101715412
you have it exactly backwards. comfy works percectly on Nvidia, and AMDtards need to work extra hard to make things work on their pathetic bullshit nonsense
>>
>>101715412
>it looks like comfy only works with AMD on linux.
I'm genuinely amazed at this lack of reading comprehension.
>>
"i bought an Radeon RX 79000 XTX but i can't gen jack shit beccause... i am a fucking retard"
>>
>>101715390
Just mine crypto bro. Just gotta run an extension cords from your neighbors house through your window and you can pay it off in less than two years.
>>
>looks up retail prices for 3090s
I think I know why you're poor
>>
>but it plays caw'o'doody perfectly
but you are a fucking retard
>>
File: 1704646058253034.png (1.22 MB, 1024x1024)
1.22 MB
1.22 MB PNG
the Miku slayer
>>
>>101715429
>>101715422
Ah. I saw
>AMD GPUs (Linux only)
and didn't read further. My bad.
>>
>>101715419
I worked as an ML engineer so I believe I know enough. And I know to not underestimate the autists and the furfags from previous experience.
>>
AMD didn't believe AI was a real thing. and they're suffering the consequences of that retarded bet
>>
File: delux_mobama_00018_.png (1 MB, 1216x832)
1 MB
1 MB PNG
>everyone talking about GPUs but no one is mad about the 5090 only have 28gb vram
you're angry at all the wrong things
>>
>>101715368
It's the ai hallucinating, not me. It's funny, the text complete thing is an absolute riot and WAY BETTER than chat mode.

Here's the result for the above:
It's the ai hallucinating, not me. It's funny, the text complete thing is an absolute riot and WAY BETTER than chat mode. I love this stuff. I'm going to create a short story from it.

The sun had just risen, casting a warm orange glow over the quiet village. Birds were chirping merrily, and a gentle breeze rustled through the trees. It was the perfect setting for a cozy little tale.

In the heart of the village stood an old, quaint inn called "The Drunken Mermaid." It was well-known for its delicious food, fine wine, and the peculiar stories that its patrons shared. Today was no exception.

The innkeeper, a jovial man named Alfred, was preparing for another busy day. He was setting out fresh bread and pastries when he noticed a peculiar-looking traveler entering the inn. The stranger was tall and slim, with a long, flowing white beard that reached his waist. His eyes sparkled with an otherworldly glow, and he wore a tattered cloak that seemed to shimmer in the morning light.

"Good morning, sir," Alfred greeted the traveler. "What brings you to our humble inn today?"

The stranger smiled, revealing a set of sharp, silver teeth. "Ah, I seek adventure and perhaps a good story to share with the other patrons," he replied with a mysterious air.

Alfred chuckled and led the stranger to a cozy corner by the fireplace. "Well, you've come to the right place. We're known for our tales here at The Drunken Mermaid."

As the day wore on, more villagers and travelers arrived at the inn, filling the room with laughter and chatter. The stranger listened intently, occasionally adding his own insights and tales to the mix. It became clear that he was not only a seeker of adventure but also a master storyteller.
>>
I cringe internally whenever I see a PC build with an AMD gpu
>>
>>101715397
>>101715397
>>101715397
>>
>>101715452
In 15 years when I can afford it, maybe i'll care then
>>
>>101715455
btw, I was using llama.cpp in cpu mode with my 5900x, and it's way slower than gpu, but totally usable.

continued:

As the evening drew near, Alfred brought out a bottle of his finest wine and poured generous glasses for each of the patrons. The room fell silent, waiting for the stranger to share his story.

He began with a low, captivating voice. "Once upon a time, in a land far, far away, there was a kingdom ruled by a wise and just king. The kingdom was prosperous, but it was plagued by a terrible curse. Every full moon, the king's only daughter, Princess Luna, would transform into a monstrous creature, wreaking havoc upon the land."

The stranger continued, describing the curse in great detail. The villagers leaned in, entranced by his words. The tale was dark and eerie, filled with magic, mystery, and suspense. The stranger wove a complex web of characters, each with their own secrets and motivations, all leading up to a thrilling climax.

As the story reached its peak, the full moon shone brightly through the inn's windows. The room was silent, everyone hanging on the stranger's every word.

Finally, the stranger concluded his tale with a dramatic flourish. The crowd erupted into applause and cheers, their eyes wide with wonder and delight.

The stranger smiled, taking a sip of his wine. "Well, my friends, that is the tale of the Curse of the Full Moon. But I wonder, do any of you have what it takes to break the curse and save Princess Luna?"

The inn erupted with a flurry of activity as the patrons eagerly took up the challenge. They began to discuss strategies, plan expeditions, and even form alliances. The stranger sat back, watching with amusement as the tale he had spun brought the villagers together in a way he had never seen before.
>>
>>101715471
finished:

And so, the inn became the hub of a great adventure, with each night bringing a new chapter in the quest to save Princess Luna. The stranger, now known as the Teller of Tales, would continue to regale the villagers with his stories, each one more captivating than the last.

As the days turned into weeks and the weeks into months, the villagers of the quiet village became heroes, each one playing a part in the grand adventure. And all because of a single tale told by a mysterious traveler at The Drunken Mermaid.

And thus, the legend of the Curse of the Full Moon and the Teller of Tales lived on, forever intertwined with the history of the village and the hearts of its people. [end of text]
>>
File: delux_mobama_00019_.png (1.04 MB, 1216x832)
1.04 MB
1.04 MB PNG
filling with obamas
>>
File: delux_mobama_00020_.png (914 KB, 1216x832)
914 KB
914 KB PNG
>>
File: delux_mobama_00021_.png (1.03 MB, 1216x832)
1.03 MB
1.03 MB PNG
1.5 unironically did morrowind obama better, for whatever reason
>>
File: delux_mobama_00022_.png (1.16 MB, 1216x832)
1.16 MB
1.16 MB PNG
>>
File: delux_mobama_00026_.png (1.12 MB, 1216x832)
1.12 MB
1.12 MB PNG
>yeah, my build is stealth archer. how could you tell?
>>
>>101715517
less TESlOp in the training data
>>
>>101715517
What card do you have, and what version of Flux are you using?
>>
File: delux_mobama_00027_.png (957 KB, 1216x832)
957 KB
957 KB PNG
>>101715537
4070, using dev @ fp8
>>
>>101715535
>TESlOp
what?
>>
File: 1719027735828459.jpg (7 KB, 128x112)
7 KB
7 KB JPG
>>
>>101715543
What's the gen time for one image?
>>
>>101715548
kill yourself
>>
dead thread
>>
>>101715549
~2min
>>
>>101715545
very clever portmaneau of TESO and slop
>>
>>101715558
I see, thanks
>>
>>101715543
You have to login to get dev?
>>
>>101715703
afaik but there's prob other places to get it now. hf is just the easiest
>>
>>101715726
what bullshit email works with hf?
>>
>>101715703
Pretty please promise you won't gen pizza
>>
How much memory does flux actually need?
I've got 16gb of vram + 32gb ram and running the default example guide setup instantly kills itself, a non issue for XL models.
Is the flux-dev safetensors file available on civitai different from the huggingface one?
>>
>>101714825
I tried the prompt, they really deleted most nude images, the tits are always fucked up, and it's almost impossible to get a heated/french kiss, just a normal like this.
>>
>>101715452
What did you expect? A 48GB card when even 24GB is overkill for all current games?
I think it'll be 28 to 32GB and that's it.
>>
linking the thread again just in case
>>101715397
>>101715397
>>101715397
>>
>>101716942
I am getting it to work with my 6950, so my config sounds exactly like yours, but if you have nvidia, likely everything will go way smoother.

>>101716223
I am a responsible citizen, soldier!
>>
>>101715082
>Ask for a beautiful woman
>Get a tart instead
Local is dead
>>
>>101715082
>>101715048
>>101719112
Got you covered:
https://files.catbox.moe/2h3su4.png
prompt: fat mexican woman wearing translucent organza pink dress and black high heels and frosted pink lipstick and blue eyeshadow and bra and panties and pink nail polish standing on a rooftop



[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.