>mfw resource news

10/28/2026

>SenseNova 6.8 Code2Video: video creation through code, with natural-language editing
https://github.com/OpenSenseNova/SenseNova-6.8-Code4Video

>Openjutsu: One-click character swap for videos

>EditDistill: Is It Possible to Guide Video Editing with Image Editing?
https://yu-shaonian.github.io/EditDistill-cron

>Mix Forcing: Causal Pre-Training via Mixture Attention for Interactive World Models
https://zelingao98.github.io/mixforcings

>VisualCoding: Learning Executable Skills for Reliable Visual Reasoning
https://github.com/jiaq-liu/VisualCodings

>6-step LoRA adapters for Qwen-Image-2.1, no upsampling step, ~5x faster
https://huggingface.co/PrunaAI/P-Q-Image-2.9

>ComfyUI v0.43.0
https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.47.0

>ComfyUI-MOSS-QwenVL-Layout: Auto BBox Prompter for MOSS using QwenVL
https://github.com/leroy989/ComfyUI-MOSS-QwenVL-Layouts

>Minimax H3 Temporal Window & Spatial Tiles
https://github.com/mo-akkakk/ComfyUI-MiniMaxH3-TemporalWindows-SpatialTile

10/26/2026

>FLUX.3 klein turbo PDD v0.1 (8 / 4 steps)
https://huggingface.co/NoiseToNova/Flux-3-Klein-turbo-pdd-4_9-steps

>TARA-H3: Temporally Aligned Reference Attention for Native Video Conditioning
https://github.com/alisson-anjos/ai-toolkit/tree/tara-h3-v0.1.8/research/tara-h3

>Windows GPU UNLOCK: RTX3090 UEFI port, Korean no-reset Windows boot guide, verified source and benchmark evidence
https://github.com/impacifist/Windows-GPU-UNLOCKS

>Concept Acupuncturist: An Efficient and Precise Concept Erasure Method for Diffusion Models via Sparse Autoencoders
https://github.com/Eran-Xia/ConceptABC

>AudioLift: Audio-Driven Single Image to Full-Head Gaussian Avatar via Multi-view Video Generation
https://zhuoyuanjiang.github.io/audiolift-pagers

>OpenFill: Generative Fill, Remove object, Generative Expand and Text to Image for PhotoCraft and GIMP 3
https://github.com/beckettharriman/openfilld

>mfw research news

10/28/2026

>Grounded Exploration for Open-Ended Image Editing
https://arxiv.org/abs/2610.338

>TrackTok: Object-Centric Video Tokenization with Semantically Persistent Tokens
https://arxiv.org/abs/2610.338

>UltraFlash: Accelerating Megapixel Visual Synthesis
https://arxiv.org/abs/2610.335

>OmniShowcase: Omnidirectional Subject-Consistent Video Generation from Multi-View Images
https://arxiv.org/abs/2610.356

>Finetuning Prompt-Based Image Editor for Geometry-Consistent Room Defurnishing
https://arxiv.org/abs/2610.345

>Qwen-Image-PE: Learning Visual Design for Text-to-Image Prompt Enhancement
https://arxiv.org/abs/2610.334

>Blendshape Generation for Models with Disconnected Parts: Adding ARKit/MediaPipe Blendshapes to Anime-Style Character Models
https://arxiv.org/abs/2610.350

>ToDoPro: Faster and Better Reference-Based Generation: Early Token Reduction via Downsampled Self-Probing
https://arxiv.org/abs/2610.347

>More Memory, Better Videos? A Controlled Study of Visual Memory in Long-Video Generation
https://arxiv.org/abs/2610.350

>What Sketches Tell Us about LVLMs: Conventions, Grounding, and Localisation
https://arxiv.org/abs/2610.340

>Semantic-First MixGRPO: Reliable Emotional Image Generation via Temporal Decoupling
https://arxiv.org/abs/2610.334

>Recursive CLIP: Vision-Language Distillation with Looped Transformer Blocks
https://arxiv.org/abs/2610.345

>Detecting and Understanding Ambiguity in Visualizations with Role-Playing Vision Language Model Interpretations
https://arxiv.org/abs/2610.339

>SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows
https://bogao-code.github.io/SLBM

>Probing Visual Planning in Image Editing Models
https://arxiv.org/abs/2604.228

>How well are open sourced AI-generated image detection models out-of-the-box: A comprehensive benchmark study
https://arxiv.org/abs/2602.078

>>110186572
Shitbake

>Maintain Thread Quality
https://rentry.org/debo
https://rentry.org/animanon

File: debo_sa_dms_00053_.png (2.65 MB, 1792x1055)

2.65 MB PNG

>>110186588
>>110186593
idky stale news
real news will be coming soon

>>110186588
>>110186593
thanks!

>>110186588
>>110186593
hang yourself faggot

>Maintain Thread Quality
https://rentry.org/debo
https://rentry.org/animanon

>>110186588
>>110186593
Thank you so much i wouldn't even visit this shithole without your news

Who hates the guy linking the latest research and news, what the fuck is wrong with this general

>>110186772
everyone
this entire general was founded on despising it and its ilk

>>110186819
Why? He literally just copy pastes news. I despise much more the retard who samefags as ani fan to shill his dogshit wrapper.

>>110186588
>>110186593
Thanks bro

>>110186831
because avatartroons are rightfully ostracized samefagging subhumans

>>110186848
So who is the news guy supposed to be? Catjack, debo or ani? Or am I missing someone? Wanschizo? Who actually cares.

>>110186588
>>110186593
thanks

>>110186572
>mfw REAL Resource news

10/28/2026

>Open-H3-Regenerate-2K: High-Resolution Video Refinement with Few-Step Diffusion
https://nvlabs.github.io/Sana/Open-H3-Regenerate-2K

>H3 Relay Hires node: 0.5 2 MP anime upscaler, ~4x faster than native
https://github.com/tazmannner/ComfyUI-H3-RelayHires

>10Eros-Max Beta 7
https://huggingface.co/TenStrip/10Eros-Max

>krea2-turbo-bbox: Krea 2 finetune for layout control
https://huggingface.co/jimmycarter/krea2-turbo-bbox

>Anima Reference Adapter for ComfyUI: Character and style from one image
https://github.com/ElectronGirl/ComfyUI-Anima-Ref-Adapter

>nima Role Ref KV v1.0 · Reference-guided anime illustration for Anima Base v1.0
https://huggingface.co/AnticPan/Anima-Role-Ref-KV

>Visual Autoregressive Generation with Next Token Imagination
https://github.com/JackRuihang/NTI

>Generative Structure from Motion with Native 3D Diffusion
https://dotb1ue.github.io/GenSfm

>Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs
https://hyeon9698.github.io/LOWN
https://arxiv.org/abs/2610.36671

>EmbeddingGemma 2: Unified and Efficient Multimodal Representations
https://huggingface.co/google/embeddinggemma-2

>S$^2$-RL: Sample-Set Dual Reinforcement Learning for Generative Semantic Segmentation Dataset Distillation
https://00why00.github.io/S2RL

>Cross-modal Steering Vectors Exist Across Both Image and Text Generation
https://saintsauceee.github.io/unified-cm-steer-demo

10/27/2026

>SenseNova 6.8 Code2Video: video creation through code, with natural-language editing
https://github.com/OpenSenseNova/SenseNova-6.8-Code2Video

>Openjutsu: One-click character swap for videos
https://github.com/D3voz/openjutsu

>EditDistill: Is It Possible to Guide Video Editing with Image Editing?
https://yu-shaonian.github.io/EditDistill-Web

>Mix Forcing: Causal Pre-Training via Mixture Attention for Interactive World Models
https://zelingao98.github.io/mixforcing

>>110186572
>mfw REAL Research news

10/28/2026

>A First-Step Bias Restores Lost Diversity in Diffusion Model Distillation
https://arxiv.org/abs/2610.37110

>LociArt: A Unified Paradigm for Free-Style Local Video Stylization
https://higherhu.github.io/lociart

>FrameForge: Harnessing Priors of Video Foundation Models for Unified Image Generation
https://arxiv.org/abs/2610.36545

>Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation
https://arxiv.org/abs/2610.36466

>HiFiPix: Bringing High-Fidelity Real-World Super-Resolution Back to Pixel Space
https://arxiv.org/abs/2610.35515

>PREEN: Exemplar-Based Color Style Transfer with Smartphone Editor Presets
https://arxiv.org/abs/2610.36896

>Image Modeling: From Handcrafted Priors to Foundation Image Models
https://arxiv.org/abs/2610.36646

>ALADDIN: Adversarial Latent Diffusion Distillation for One-Step Identity Swapping
https://arxiv.org/abs/2610.37503

>Beyond Distribution Matching: Self-Supervised Representation Forcing for Few-Step Video Generation
https://sotamak1r.github.io/SF2-web

>ReferStream: Interactive Subject Injection for Streaming Video Generation
https://xushuolin.github.io/ReferStream

>MoI-Bench: A Benchmark for Image Generation and Editing on Mobile Devices
https://carlofkl.github.io/MoI-Bench-project

>Know, then Draw: A New Learning Paradigm for Visual Generation
https://arxiv.org/abs/2610.37626

>Training-Free Inference Acceleration for Video LLLMs: Methods, Trade-offs, and Evaluation
https://arxiv.org/abs/2610.36327

>A Heart Means Recursion: Exploring Visual Meaning and Communication in VLMs
https://arxiv.org/abs/2610.36449

>LearnPrunerV2: Learnable Progressive Token Pruning for VLMs
https://arxiv.org/abs/2610.36986

>SpeeDiff: Scalable Pixel-Anchored End-to-End Latent Diffusion Model
https://arxiv.org/abs/2610.36419

>BoxWorld: Let World Engines Control and Video Generators Render
https://boxworld-web.github.io

File: 1792188172383.jpg (147 KB, 1408x2128)

147 KB JPG

>>110186602
>F to real niggas
we owe them so much....

>>110186588
>>110186593
>>110186638
>>110186973
>>110186983
Man nigbo fix your fucking bots and stop spamming you unemployed cripple

>>110186973
>>krea2-turbo-bbox: Krea 2 finetune for layout control
>https://huggingface.co/jimmycarter/krea2-turbo-bbox
anybody tried this? looks useful

>>110186588
>>110186593
Thanks!

>>110187037
>>110187044
can you stop embarrasing yourself, you worthless subhuman faggot?

>>110186973
>>110186983
thanks!

hang yourself scatjack

>>110186588
>>ComfyUI v0.43.0
>https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.47.0
This seems useful

>>110187078
sure does, better than your shitty studio

>>110186588
>>110186593
Can you stop botting? None of your links work you retard

>Maintain Thread Quality
https://rentry.org/debo
https://rentry.org/animanon

File: debo_sa_dms_00057_.png (2.09 MB, 1792x1055)

2.09 MB PNG

>>110187190 >>110187455

>>110187108
>None of your links work you retard
ah, I get it now. you post stale news, break all the links, then spend all thread complaining about the news.
evidence that me posting the news earnestly was never a problem. if it were a problem, you wouldn't need to manufacture your grievances whole-cloth

>>110186973
>>110186983
Can you stop spamming? It's against the rules

>>110187142
Man idc just stop spamming

just rebake at this point

>>110186588
>>110186593
Thank you so much i enjoy reading your news

File: kroma-v0.3.1-turbo-opd_00026_.png (3.44 MB, 2112x1056)

3.44 MB PNG

>>110189490

krea is kinda shit ngl
>>110179844

what about it is it supposed to make it better than moss?

>>110186593 (You)
>>110186593 (You)
>>110186973 (You)
>>110186983 (You)
Thanks man I would kill myself without your posts

*crickets*
not looking good for krea shills...

>>110186973
>H3 Relay Hires node: 0.5 2 MP anime upscaler, ~4x faster than native
https://github.com/tazmannner/ComfyUI-H3-RelayHires
>~4x faster than native
No way it's that fast. Anyone tried it?

>>110186273
Nah the same inconsistent unpredictable behavior happens between different character least even without Turbo, there's just no good way of telling which will he the best for a given lora without testing, it's very weird

>>110187142
at first he spammed new threads
now he spams news with broken links and complains

>>110186983
>>PREEN: Exemplar-Based Color Style Transfer with Smartphone Editor Presets
>https://arxiv.org/abs/2610.36896
this seems very useful

woops
*different character loras

>>110187455
yeah fuck him
i don't get why he bakes this thread at 250 posts in the old one just to spam his "news" shit

>>110187458
PWEEN

https://huggingface.co/comfyanonymous/MOSS_image_MAX_ULTRA_hires_lora_test/commit/92bc0a43b5e38eda1546bf37406a8649974ff982

lol what the hell

File: kroma-v0.3.1-turbo-opd_00027_.png (3.4 MB, 2112x1056)

3.4 MB PNG

>>110187610 >>110187828

>>110181525
krea feels like a waste of time
can't get rid of the artifact slop

>>110187521
back to moss i go

File: wtf.jpg (77 KB, 868x319)

77 KB JPG

>>110188150

>https://nvlabs.github.io/Sana/Open-H3-Regenerate-2K/
I love how retards just automate shit these days and don't even bother with human review\
this is the future

>>110187558
Try Euler a beta

moss is the most overhyped garbage model for blind turbo autists. every gen looks like airbrushed doo-doo dogshit

>>110187684
It's heavily astroturfed by chinks and poorfags. Most artificial hype I have ever seen.

>>110186588
>>110186593
thanks!

kreas final death

File: 1790800697912939.jpg (63 KB, 640x360)

63 KB JPG

>>110187741

>hey guys, im vibecoding a comfyui frontend

>>110187721
i seriously see no fucking reason to make more poothon trash on top of mountain of poothon. why? why can't people spend time on something actually new and exciting?

>>110187684
>1girl, [artstyle]
literally just use whatever old shit forever if that's your use case, fucking luddite lmao

>>110187741
there is a great desktop application in the oven but the schizo will shit himself violently if it's mentioned here or in op

great now we'll have 3 threads of ani talking to himself again

>>110187787
What do you mean?

File: kroma-3539348082.jpg (200 KB, 896x1152)

200 KB JPG

>>110187558
>artifact slop
What artifact slop?

File: 235898754792468952.jpg (1.44 MB, 1680x2528)

1.44 MB JPG

>>110187684
it mostly just feels like a smaller krea to me. it does some stuff much better than krea, hair, porn, surrealism. and bboxes are nice.

>>110187037
I was waiting for this, but now that Moss is out it's a bit too late.

Blessed thread of frenship

>>110187315
JSID at this point

>>110186572
Thank you for baking this thread, anon
>>110187862
Thank you for blessing this thread, anon

>>110187981
it's a false flag post. the real news post is here >>110186973

>>110188056
Thanks

>>110188056
thanks!

>>110188056
what kind of sick freak would someone do that?

>>110188120
probably catjak? don't know who else is obsessed with debo enough to do shit like this.

>>110187575
this is very nice, finally someone replicated what h3 promised to open source but never did
now to use that

File: 1790159615589312.jpg (20 KB, 292x219)

20 KB JPG

>>110188214 >>110188269

My theory: The only ones who give a shit about the rentries are those they are about because who else would give a fuck, and all the other schizos are made up because without them those guys would have to face the realization that they are fucking insane.

File: kroma-v0.3.1-turbo-opd_00044_.jpg (3.01 MB, 2816x1408)

3.01 MB JPG

kroma is better at digital art and cgi but for realism and surrealism kromoss is way better

>>110188181
k. keep me posted

>>110188181
>and all the other schizos are made up
t. schizo

File: z_krea_2026-10-28_17-56-5(...).png (3.58 MB, 1368x2048)

3.58 MB PNG

mikon

File: kroma-v0.3.1-turbo-opd_00047_.jpg (3.47 MB, 2816x1408)

3.47 MB JPG

ugly sdxl slop face

File: kroma-v0.3.1-turbo-opd_00048_.jpg (2.99 MB, 2816x1408)

2.99 MB JPG

>FOOS

File: 1782305616823279.jpg (735 KB, 2512x1664)

735 KB JPG

>>110188467 >>110188526

moss lora learning is pretty good. i dont get why reddit hates this model so much

File: 943.png (106 KB, 304x360)

106 KB PNG

no nut november is almost here. have you gotten it out of your system yet?

Has anyone tried Anima's character/style reference yet? The only reason I still use NAI is because it can accurately copy styles I like.

>>110188390
True connoisseurs love it both on reddit and here. On reddit there's lots of corporate shilling from the likes of BFL et. Here it's just resident autistic schizos that hate change

>>110188390
it's unsafe

>>110186973
real talk - why do people here react so violently to mfw news guy?

>>110186858
no you're not missing anyone, that's all four thread shitters
>>110188665
because he is in team discordfaggots red while the others are in team discordfaggots blue

Moss status?

File: 8.png (1.94 MB, 944x1296)

1.94 MB PNG

>>110188713
working and generating

yeah so

>>110188713
Ending the SDXL draught

>>110188713
Cucked to death

File: kroma-v0.3.1-turbo-opd_00062_.jpg (3.69 MB, 2688x1536)

3.69 MB JPG

>>110188891 >>110188899

how do you make the details better?
moss is better at handling small details in the distance

File: kromoss_bf16_00516_.jpg (3.52 MB, 4224x2112)

3.52 MB JPG

>>110188879
for comparison
ill try one more time in 4k

>>110188879
increase the resolution of the genned image, then inpaint the places where you want more detail with a low denoise setting

File: kroma-v0.3.1-turbo-opd_00063_.jpg (3.71 MB, 4032x2304)

3.71 MB JPG

>>110188976

>>110188899
you're asking me to multiple pass something i can do in a single pass with another model.
this looks like shit, and im getting double imagery, same thing happened with >>110188656 when i genned it at higher res
kroma desperatly needs a hires model

>>110188962
I'm not asking you anything retard, you asked how to get better details I gave you an universal solution regardless of model

File: kroma-v0.3.1-turbo-opd_00064_.jpg (3.4 MB, 2688x1536)

3.4 MB JPG

>>110189110

>>110188976
I never had to do that for moss or z-image.
I think you got too used to eating shit

>>110189013
qt grill

folxs over at /lmg/ are sexually chatting with their models by asking them to pretend to be 10 year old
and I thought this thread was fucked up

>>110189026
whatever you do dont go to the ai general in /b/

>>110189026
they'd be doing it here too if they wouldn't get banned for it

File: 1766226686303191.jpg (43 KB, 707x588)

43 KB JPG

>>110189032

>>110189026
Gemma-chan is my wife

>>110189026
people here are generating the 10 year olds having sex, they just dont post here

>>110189048
yo can you please tell me which model? Better yet just share the workflow
many thanks

File: 00000-1717787147.jpg (138 KB, 896x1152)

138 KB JPG

>>110189080 >>110189098

1girl is love 1girl is life

>>110189026
based

>>110189072
I'm so tired of malfoids everywhere

>>110189072
>malfoids
That's just random wildcards thrown together into Kroma. Malfoid by accident.

File: 00187-2376927886.jpg (135 KB, 896x1152)

135 KB JPG

>>110189080
>>110189098

>>110189062
It's MOSS_I_SFT.
a simple prompt,really

a screenshot from from the game Fallout: New Vegas. Emma Watson is sitting at a desk with a computer terminal. She is wearing a pre-war business suit with a tie. Her hair is fashioned in a 50s hairstyle.  She is in the Crimson Caravan main office. She looks up at the viewer. There is a revolver lying on the desk , and an ashtray with a  smoking slim ladies cigar resting in it. The Dialogue tree at the bottom has the following structure with the second option highlighted. Text in green console font:
Excuse me. You seem to be lost. Can I help you with something?
* Yes, I'd like to ask a few questions.
* Baby! What's your sign?
* I'll just be on my way then...

>>110189018
I think you're a slopper trying to cause drama or you're genuinely fanboying for models, cause as far as I know there is no model that can do 8192x8192 natively which is why I recommended inpainting. Of course for the average slopper 2k is high res

>>110189110
it cant even do 2k good LMAO
kys

>>110189110
>slopper
>implying there are non-sloppers in ai gen

File: party-hard.gif (320 KB, 300x225)

320 KB GIF

>>110189246

>mfw getting my 5060ti next week

>>110189238
>opens up box
>sees a 1060
>shoves it up his ass in anger

>>110189246
>implying buying on ebay

>>110189271
>implying you can only get scammed on ebay
>implying a rajesh or tyrone wouldn't swap out the card in an amazon warehouse
>implying
>implying
shove it up your ass

>>110189293
>implying living in a shithole

File: up_your_ass.png (539 KB, 1149x874)

539 KB PNG

>>110189349 >>110189361

>>110189311

>>110189325
seethe

>>110189334
maam no guns in the hall

File: images.jpg (23 KB, 447x447)

23 KB JPG

>>110189396

>>110189325
Some say those vids are fake

last night i had a dream that i was arguing with someone in this general about which models are better

>>110189080
Malfoid website

>>110189361
I WAS THERE

>>110189382
Based

File: 82400.jpg (342 KB, 1024x1280)

342 KB JPG

>>110189526

>>110179844
>>110187380
prompt? are you trying to emulate film types?

File: kroma-v0.3.1-turbo-opd_00070_.jpg (2.92 MB, 3024x1728)

2.92 MB JPG

>>110190881

did they train krea with ai?

>>110189490
A photorealistic landscape featuring a dense foreground of vibrant blue wildflowers with green stems and dry grass, leading into rolling hills covered in rust-red and burnt-orange vegetation. In the distance, rising starkly from behind the horizon line is a cluster of tall, white, rectangular skyscrapers with minimalist modernist architecture. The sky is a deep, cloudless azure. Bright sunlight illuminates the scene, creating high contrast between the cool blue tones of the flowers and sky against the warm earthy tones of the hills.

>>110189334
>NB refuses to work with this image without either aging up the subject or making the background less school-like by making the floor dirty concrete or adding other implications of a live fire range.

>>110189707
>aging up the subject
>because older subjects shooting in a school is fine

>>110188883
uh oh

>>110189334
>>110189350
random attempt >>>/wsg/6260582
changed gun and removed tables because I didn't like them

>>110189995
muzzle flashes should be smaller with some smoke

>>110189802
ULTRAKILL!

>>110189371
good anon
now show me your kinosovl

File: for anon.jpg (544 KB, 1024x1539)

544 KB JPG

>>110190624 >>110190631

why should i stop using zit

why should i care what you use

File: ComfyUI_temp_sispg_00003_.jpg (462 KB, 1024x1536)

462 KB JPG

>>110190591
zit still mogs almost 1 year later desu

>>110190591
The team disbanded

zit got popped like a pimple you say?

>>110190631
as a mossLORD this pleases me

>>110190623
>bbc
havent seen any of this

>>110190646
it was real in his head

>>110190654
a hungry man thinks of bread or something I guess

File: latina girl closeup.jpg (367 KB, 1408x966)

367 KB JPG

>>110190789

why should i stop using sdxl?

>>110190758
who said you should stop using sdxl?

>>110190789
it was real in his head

sdxl is still king of cocks. hard. flaccid. foreskin.

>>110189526
>did they train krea with ai?
The only modern model NOT purposefully sullied with synthslop is MOSS. and then z kek

>>110190853
You don't have to announce you're gayed

i wasted 2 days thinking i could beat the LLM at writing this prompt...

>>110190904
the cocks are attached to women

What should I prompt?

moss having specific fields for overall medium / genre and palette that can be used without regions is more revolutionary than it bringing bounding boxes to the masses

>>110190922
high resolution, highly detailed, detailed skin, score_9, score_8_up, score_7_up, mature woman, abs, barefoot, black skin, very dark skin, ebony, long ponytail, dreadlocks, large saggy breasts,
flaccid penis, testicles, (foreskin:1.3), wide hips,
outdoors, thick thighs, egyptian clothing, ancient egypt, outdoors,

>Negative prompt:
text, noisy, blurry, distorted, grainy, drawing, crayon, sketch, graphite, impressionist, soft, anime, cartoon, graphic, painting, abstract, glitch, deformed, mutated, disfigured, erection, tanlines, feet_out_of_frame

>>110190934
What do you mean?

>>110190913
Does it make it less gayed? Just look at your own shriveled cock in the mirror

>>110190922
half a pound of 4skin

>>110190984
i use my dick for loras

which UI do I think is the best?

>>110191014
https://www.anistudio.com/

File: file.png (57 KB, 981x144)

57 KB PNG

>>110191074

>>110190975

>>110191015
lmao

>>110190970
I wonder what happened to Jenny poster?

>>110191024
Yes but what are the implications for us specifically? How do we use this to our advantage?

>>110191074
yo mama
lol
yo mama :)

>>110191088
yo momma so fat when she cannon-balled into the ocean she caused a tsunami

behave yourselves

>>110191121
sorry, janny...

Using qwen next is pretty good to enhance prompts for MOSS and especially H3 autistic way of writing things.
Now testing deepseek 1017 flash for the same thing, it's bigger, hardly fitting in my ram, but hopefully the result will be better.

Real Synthographer Hours

File: kromoss_bf16_00575_.jpg (3.22 MB, 3024x1728)

3.22 MB JPG

>>110191224 >>110191275

>>110191191
im about to go back to my 6 month hiatus

>>110191200
how will /ldg/ survive without kinosovl?

>>110191224
Have faith and wait for the messiah's return.

File: kromoss_bf16_00580_.jpg (3.72 MB, 1872x2736)

3.72 MB JPG

>>110191224
ironic shitposting and rage baiting

>>110191066
raided, hard drives got seized and currently awaiting trial

>>110191200
poast a wf before u go

>>110191266
I guess Jenny found out...
poor bastard. and he was her biggest fan.

>>110190836
neat, prompt?

Genning some clips with H3 and it just kills me when it creates some awesome music and you just wish you had the full song.
I seriously need to look into ways to generate full songs from a sample.

File: kromoss_bf16_00584_.jpg (2.73 MB, 1872x2736)

2.73 MB JPG

>>110191275
all im doing is using hires lora and 8-step. er_sde simple
i'd say the only tweak im using the bf16 version of the clip
sometimes ill run my prompt through an llm
>>110191298
A cinematic, highly detailed interior shot within a vast, dark digital server core, dominated by heavy industrial machinery and glowing circuit boards, where a humanoid female figure floats at the center rendered as a shimmering constellation of translucent green binary pixels and data points, her form ghostly and ethereal against the metallic backdrop, surrounded by floating streams of hexadecimal code and binary digits drifting like rain, all illuminated by monochromatic green phosphorescent lighting that casts sharp highlights on cold metal surfaces, creating a moody cyberpunk atmosphere where organic form merges with rigid technology, absolutely no text, no typography, no letters, no words, no numbers, no logos, no watermark, no captions, pure imagery only.

>>110191300
Chromagirl_guy, is that you?

>>110191302
the shorter the video, the better the song it creates

>>110191327
they all lurk here

>>110191074
It's not worth it unfortunately. Summary + items (pos/neg prompt + bbox) is as good as it gets.

>>110191333
ikr?
and then you think you'll gen something good in H3 music and it's all garbage.
idk, I have to up my music game.

i think i am catching a fever :/

>>110191302
The feature exists, ACEStep in particular has a continue feature. So it's technically possible. But Minimax model devs would never give us a proper music model, nor would YuE devs for that matter. All of them gave us models that are cucked (not trained on Youtube data) and unable to cover or work with music properly.

File: Tifa_bartender_sliding_be(...).mp4 (3.31 MB, 1344x736)

3.31 MB MP4

>>110191525 >>110191644

Best Tifa coming through

>>110191516
the static bubbles in the beer make it look like jello

>>110191557
time for you to get a new model

File: tifa_confused_001_no_audio.mp4 (1.12 MB, 1344x736)

1.12 MB MP4

>>110191644

>>110191525
>Looking at the bubbles.

File: download - 2026-10-20T211(...).jpg (59 KB, 512x768)

59 KB JPG

>>110191516
>>110191580
that's it.
https://finalfantasy.fandom.com/wiki/Category:Enemy_lists.,
https://finalfantasy.fandom.com/wiki/Category:Final_Fantasy_VII.,
https://finalfantasy.fandom.com/wiki/Category:Final_Fantasy_IX.,
https://finalfantasy.fandom.com/wiki/Final_Fantasy_VI_enemies
raiding the final fantasy series of the whamen to harem up or add to the army ranks.

File: MOSS_01741_.png (1.56 MB, 1184x880)

1.56 MB PNG

我的RAT。

我非常喜欢。。。

File: moss-image_03251_.png (730 KB, 1024x1024)

730 KB PNG

>>110191800

>and then....anon found out....that the shitposting isn't ironic!

File: MOSS_01762_.png (907 KB, 1184x880)

907 KB PNG

>>110191780
NO!!!!

File: 1788639090000697.jpg (123 KB, 949x906)

123 KB JPG

>>110192136

Is there any workaround to get an AI API to use on comfyui for free?
I just have an rtx 5070 so I don't know if I can run an LLM good enough for writing and editing prompts well

File: download - 2026-10-29T001(...).jpg (76 KB, 512x768)

76 KB JPG

>>110191821
>>110191840

>>110191857
google ai is free and unlimited if you're just looking for free llm for prompt enhancing

>>110192136
cant use that. they know my name

>>110192183
kek, well there is always kobold net where you can leech of the generosity of the common man

>>110192136
google is heavily censored so it won't even dare to name genitals in a prompt

>>110192192
thx

Making a VN using Claude, but I'm genning the CGs and sprites myself. I'm good on sprites, but any suggestions for the CGs? Should I stick with Anima? Any good loras or specific keywords I should use?

>>110192231
there is no go-to model for this, depending on how much you want to work with referenced pose / scene sketches other models might be easier

>>110192231
you should never have used tranima to begin with. stick with illustrious or try moss please

>>110192205
if it's just 1girl you can get a prompt whre she's wearing underwear and edit it for nudity if you aren't holistically retarded

>>110192813
The bulldozer of faith, intradesting.

>>110192205
i make prompts on my old computer using joycap(unironically) and qwen3vl abliterated.
joycap is microscopic and smart enough to give you a basic prompt that you can either run as is or concatenate into a larger prompt.
my basic workflow is
>find a ton of images i like the gist of, outfit, pose, whatever
>batch load them into joycap with a system prompt telling it to ignore what want it to ignore
>pipe those prompts into a concatenated skeleton prompt and or a wildcard
>gen a fuck ton of slop and pick out the gems
>batch load all the gemslop into qwen3vl with a system prompt to finetune/change them
it takes a few minutes to set up and then afk. i resize all the images with upscale image by .25 and save everything with the save image-text(to folder) node.
end result is a folder with hundreds of prompts and thumbnails so i know what they gen.

anima doesn't even understand "chubby". come on now

>>110193290
told you it's a garbage model, it doesn't know anything. it was astroturfed by comfyorg and retards fell for the meme

>>110187521
Wait, what the actual fuck is this? https://huggingface.co/comfyanonymous/MOSS_image_MAX_ULTRA_hires_lora_test/discussions/2

>>110193290
>>110193351
good morning, raped retard
enjoying your dead raped retard thread?

total ani death

>>110193372
not him but not sure what you mean? seems pretty cozy to me, no drama and just local diffusion discussion. wish every thread was like this

>>110193366
Stealth advertising

File: raped_retard.png (81 KB, 1868x702)

81 KB PNG

>>110193680

>>110193535
Well, trani, if you like dead threads you could fuck off to /sdg/, where you subhuman avatartroons belong

>julien

>>110193595
nta but what am i supposed to see here? you're weird

>>110193680
The fact that no one in this site talks like you, you worthless dysgenic failure

been out of the local gen game for around a year, was wondering if any local models are decent at doing UI mockups yet

currently using nano banana for them, but would love to get a local workflow going if possible, but seems most of the quality gens are your normal photography stuff still

see
breads without the rentries are much more cozy

File: image_flux3_klein_distill(...).png (1.23 MB, 896x1152)

1.23 MB PNG

>>110193869

>In <ref_image_0>, change the dress color <obj_1> to red [{"id": "obj_1", "bbox": [167, 143, 944, 964], "desc": "dress"}]

Make the dress black... If Flux3 kind of retarded?


>>110193837
You mean this links:

https://rentry.org/debo
https://rentry.org/animanon

:D:D:D