Stable Diffusion Tutorials

Generate PROMPTS like a PRO 😹 👇 #stablediffusion #aipromptgenerator 🔖Bookmark (save) for future reference.... 🩷 Like if you loved it.... 🖊️Comment/Suggestions if you are getting any error....

Ltx 2.3 😃Greenscreen Avatar Ic lora vertical This LoRA generates perfect green screen avatars for TikTok, Shorts, and Reels -Optimized for vertical formats 9:16 -Clean green backgrounds for easy chroma keying -Expressive avatar generations HF repo:👇 huggingface.co/OmerHagage/ltx…

Implemented😃 NegPip on the Z-image series. Neg pipe-allowing for the use of prompts with negative effects within regular prompts and prompts with positive effects within negative prompts github + workflow:👇 github.com/BigStationW/Co…

Left is Hidream O1 Base 30 steps 4 CFG. Right is a finetune I am working on, only 4 steps, no CFG. Back story, I have been finetuning 2 siglip2 naflex models this past week on datasets I have been preparing for a few months. 1 is a full image (not cropped) person identity model. The second is a style model. Doing an arcface type loss on both. Since it is naflex, they can take in varying sizes and aspect ratios. I trained them on up to 1MP images. I tested using the ID model to train a ZIT (no training adapter) lora on myself using only the new embedding output as a loss, no flow matching loss. It worked somewhat, though wasn't perfect, but more importantly, it didn't break down the turbo distillation at all. Which means, you would be able to distill a model doing something similar. So I started with just siglip2 so400 naflex training on Hidream O1. It was working, but was overfitting to the vision encoder. So I added clip-g as well. It got better. So I started multi-scaling into siglip2, even better. So I added my style model and my new id model. And it started cleaning up extremely fast. Bear in mind, I am doing all of this with a batch size of 2 at 512 on a 5090, so, even though it seems like a lot of vision compute, it is within reason. It is continuing to clean up, but now I think I need to finetune more vision encoders for different knowledge. The model will learn to trick the vision encoders, but each one added makes it harder to trick. The naflex finetunes allow you to recover fine detail on large images. Each one acts as its own expert critic focusing only on their own task. The more you have, the more knowledge you can transfer from the images to the model. Which is important, because the model never receives a loss on the real images, only the single vector image embeddings. I will keep you posted on progress.

@blendthe_rules yup... AsymFLUX.2 Klein Sampler node can do Text-to-image and image-to-image gen

@SD_Tutorial Does this support editing?

FLUX.2 Klein+AsymFlow with no VAE. -builds hyper-realistic images directly in pixel space rather than compressed latents -Sharper textures and superior fidelity40% faster -Low-rank noise parameterization solves high-dimensional -Comfy support incoming 👇 hanshengchen.com/asymflow/

Comparison 😃🧐 Anima_pv2 vs Anima_pv3 vs Animabase_v1 detailed post:🤗🤓 reddit.com/r/StableDiffus…

LTX Director 🎬 📹Complete Timeline Editor For LTX 2.3 😃Custom node for ComfyUI -Prompt Relay enable -Training Free Integration -Wan & LTX Supported -Functional Timeline Editor -Multi Keyframe Support -Custom Audio Support -Builtin T2V & I2V 👇github: github.com/WhatDreamsCost…

FLUX Klein Base 9B ☺️ SKIN Lora one of the best skin LoRA’s available right now -no more plastic skin -Miniscule improvements -V2 version more improved than v1 Civit ai:👇 civitai.red/models/2613362…

ComfyUI node pack ☺️😃for AsymFLUX.2 Klein -- pixel-space text-to-image generation -no vae github+workflow: github.com/CanFromEarth/C… Results:

Anima base v1.0 😃 -Created with collab:CircleStone Labs and Comfy -For anime concepts, characters, and styles, but is also capable of generating a wide variety of other non-photorealistic content. civitai: civitai.com/models/2458426… HF repo: huggingface.co/circlestone-la… results:👇

LTX-2.3-22b-IC-LoRA-LipDub 😃 by Lightricks HF repo👇 : huggingface.co/Lightricks/LTX…

@MrBerneker @Durpmen

@Durpmen @SD_Tutorial So many things, so little time!

Wan 2.2 Image2Video 😃+ LTX 2.3 ID-LoRA☺️ in Comfy -Generate your initial video using Wan 2.2 -LTX 2.3 automatically adds audio to the clip -LTX 2.3 seamlessly extends the video with your next scene Workflow: huggingface.co/ussaaron/workf… detailed post: reddit.com/r/StableDiffus…

LTX-2.3🎬🎦 SYSTMS-FLW IC LORA Perfect for seamless shot-to-shot transitions. HF repo:👇 huggingface.co/systms/SYSTMS-…

This is fully open source. 😃🔉🗣️The model weights derive from the LTX-2 Community License but all inference and pipeline code is MIT. All demos + article: scenema.ai/audio Model weights: huggingface.co/ScenemaAI/scen… Code + setup: github.com/ScenemaAI/scen…

every clip was made with LTX2.3 using TNG image screengrabs and this awesome lora: 👇👇 huggingface.co/bionicman69/St…

LTX 2.3 ✖ Cpdex で Music Video をフルオートで作成してみた！！ まだ、控えめな720pなのでこれは結構いけるぞ！！

Compilation of the open-source LoRAs for LTX 2.3 - released in May at one place: 👇 reddit.com/r/StableDiffus…

Meet Z-Anime by @seesee. Built on the powerful Z-Image Base architecture, this model brings flagship-level diversity and precise prompt control to anime generation. It inherits full support for complex negative prompts and extreme customizability. Ready to push your anime workflows to the next level? Grab the weights on Hugging Face:👇 huggingface.co/SeeSee21/Z-Ani…