Видео

Unfortunately a video model has yet to be trained to work with Stable Diffusion 3.5. More recent video models that are unrelated to Stable diffusion but work in comfy UI are CogVideoX and Mochi. Both require a lot of VRAM or Patience.

I am sorry to hear that, what UI and Model are you using?

Es lohnt sich!

ComfyUI bekommt bald eine übersichtlichere UI. Dann wird es etwas einfacher. Sobald ich in die Beta kann mache ich ein Video dazu.

@@AIKnowledge2Go cool. solange bleibe ich bei ner mischung aus xl und 1.5

Yeah me too. I like flux but somehow SD feels like home. Maybe its the broken hands 😂 But there is a lora for this on Civitai. Haven't tested it yet though.

Flux is better...

Thanks for joining the community! I’m excited to have you here!

Hi you need: images that represent the clothing you want to train on, including different angles, poses, and lighting conditions. If possible, use images with simple or uniform backgrounds to ensure the model focuses on the clothing rather than complex scenery. Remove blurry, low-quality, or overexposed images. Ensure that the dataset includes various body poses, especially if the clothing involves different parts like sleeves, collars, or drapes

If your outputs start to look too much like specific training images or capture excessive detail that wasn’t intended.Over-trained models might struggle to respond to prompts that deviate from the training data. If the model produces outputs that are too generic, vague, or don’t look like the specific character in your dataset, it needs more training.

This should be visible by default, in forge UI, which version are you using?

Thanks, to be honest i haven't tested it with pony yet. What happens when you render the same image with the regional prompter turned off while using the same seed? If you get the same image than pony may be the culprit here. Have fun testing.

Yes! But you need the according base model. For flux I only tested nf4 flux v2 version of flux and the "normal" flux dev. GGUF is untested.

Sorry for the late reply. What happens when you open a cmd window and type "git --version" it should say something like "git version 2.45.2.windows.1" if not then go here and download and install the latest. git-scm.com/downloads Hope that helps.

Danke, ja den Akzent kann ich nicht verbergen. 😂 Ich bilde mir ein in neueren Video ist es besser geworden😂

It's great that you figured it out on your own. I wouldn't wait for SD3. I want to love it but even stability AI seems to have lost interest in fixing it. News from 2 Month ago was that they want to release a 3.1 version in the upcomming weeks. I am still waiting. Have you tried flux yet? If not then you should. Its everything SD3 promised to be.

Thanks for the correction

I just can’t hide it 😂 You’ve got a good ear!

This seems really dumb to use renders for this.

You are absolutely right. I will make an "advanced" video of this training guide and fixing hands will be on of the talking points.

Youre wrong. Yiu can always generate images of your character using the "perfect hands" lora.

Its absolutely great to use ComfyUI for this. I saw a workflow once on civitai that created Multiple different images for Sytle Loras. Maybe I will make a video about XYZ Plot in ComfyUI. Unfortunately its not so straight forward as i thought.

Sorry for that, i had the same link twice in my description. Is fixed now. Thanks for pointing that out.: stability.ai/news/stable-diffusion-3-research-paper

SD 1.5 definitely and i think SDXL as well. You can do the same or even better with antimate diff.

I am sorry to hear that you run into trouble. Using Flux is really hard with 6GB VRAM. Have you tried using the GGUF models of flux?

Step 3: Enjoy your virtual dates without the awkward small talk! 😂

This depends on if you set a1111 to inpaint "whole picture" or "only masked" with whole picture you might want to set it to 1x. With "only masked" the inpainted area will be inpainted with the same resolution as the rest of the picture. This takes a) a long time and b) the Image cannot hold this information because of the total resolution of the image. An image with 10.000 pixels cant hold 10.001 pixels. You might want to check the newer version of this video: ruclips.net/video/wyDRHRuHbAU/видео.html

That’s a great point! It would definitely make things easier if there was a streamlined install program. you might want to give this video a try where i introduce stabillity matrix. This won't get rid of all the annoying stuff but you can install a lot of UI's with two clicks and it holds all your models, loras and created images. Its not perfect but it makes things a lot easier. ruclips.net/video/bwPk-NXggp0/видео.html

​@@AIKnowledge2Go thanks for the reply but I am not psychologically at a point where I want to do much work installing and setting up some AI image stuff and spending gobs of time trying to learn it. I did want to just check into it some and see if I could get some understanding of what's going on. I think $10/mo for some midjourney images would be okay if I was making more than $10/mo off of those images. But i'm not and I just wanted really to try some of these websites out but I did look at night cafe but I just prob gave up on that one too soon but I was wanting to make some landscapes with old mansions and you know that dark academia/fantasy look. Add some light animation after the fact with maybe multiple layers in video editor etc. Anyway.. thx for the info and explanations!!!

Hi, yeah usually newer is better in terms of checkpoints. Some Checkpoint creators pushing out new versions on a daily basis. When i came up with the idea for the video 11 was the latest and i don't change version while working on a project. This is why its still version 11.

@@AIKnowledge2Go danke dir für die schnelle Antwort :D

Du darfst... zum Zeitpunkt als ich das Video aufgenommen habe, war ich noch nicht so überzeugt von Fooocus. Tatsächlich ist eins meiner kommenden Projekte ein Inpaint tutorial wo ich Fooocus vorstellen werde

You are welcome!

Thanks for your feedback, if you point me to where the problem lies, then I may be able to improve on it in the future.

Hi danke für das Feedback. ich habe tatsächlich Pläne zu sowohl inpainting und auch outpainting neue videos zu erstellen. Leider kann ich noch keinen genauen Termin dazu nennen.

No, its a completely different installation. Your A1111 remains untouched.

I'm catching up to you. My next album cover art will be 3 half naked girls, swords firmly held with their feet .

Glad you enjoyed it!

Yes great explanation thanks, I've been waiting to create a lora and have had issues with tryint to create the captions and data set. I'm interested in training a style

You can train both in a single lora, this requires a more delicate image selection and the rendering of a lot more images. If the body has to be of nude form, then good luck. You may want to use a textured 3D Model for this and render these Images in a 3D Environment, like Blender or Unreal Engine 5. Hm... actually not a bad idea for a new video... *taking notes*. You also could take real photos, but make sure you have written consent from the person to use them.

@@AIKnowledge2Go that kind if video would be really sick man

Thanks for you feedback. Since most of of my viewers have an RTX 2080 or lower ( i did several polls on this). i decided against local training. I am using the Kohya method in Civitai online trainer. This can be applied 1 - 1 to local kohya and should yield similar results.

Hi. I installed it almost like you said. I chose standalone and now It starts pytorch and wants to use CUDA and I only have 2 AMD GPUs but it does not show the 2nd GPU. I might have ZLUDA installed somewhere too but I chose directML so I don't know why I'm getting that error. Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)] Version: f2.0.1v1.10.1-previous-613-gf10214e3 Commit hash: f10214e358e730b495ce7bfb0edc4846caecf1b2 Traceback (most recent call last): File "...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\launch.py", line 54, in <module> main() File "...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\launch.py", line 42, in main prepare_environment() File "...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\modules\launch_utils.py", line 436, in prepare_environment raise RuntimeError( RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version: github.com/lllyasviel/stable-diffusion-webui-forge/releases/tag/latest I am trying to run it without pytorch CUDA testing. It seems to do something. The web UI is starting: Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)] Version: f2.0.1v1.10.1-previous-613-gf10214e3 Commit hash: f10214e358e730b495ce7bfb0edc4846caecf1b2 Installing clip Cloning assets into ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\stable-diffusion-webui-assets... Cloning into '...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\stable-diffusion-webui-assets'... Cloning huggingface_guess into ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\huggingface_guess... Cloning into '...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\huggingface_guess'... Cloning google_blockly into ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\google_blockly_prototypes... Cloning into '...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\google_blockly_prototypes'... Cloning BLIP into ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\BLIP... Cloning into '...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge epositories\BLIP'... ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\extensions-builtin\forge_legacy_preprocessors\install.py:2: DeprecationWarning: pkg_resources is deprecated as an API. See setuptools.pypa.io/en/latest/pkg_resources.html import pkg_resources Installing forge_legacy_preprocessor requirement: fvcore Installing forge_legacy_preprocessor requirement: mediapipe Installing forge_legacy_preprocessor requirement: onnxruntime Installing forge_legacy_preprocessor requirement: svglib Installing forge_legacy_preprocessor requirement: insightface Installing forge_legacy_preprocessor requirement: handrefinerportable Installing forge_legacy_preprocessor requirement: depth_anything Installing forge_legacy_preprocessor requirement: depth_anything_v2 ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\extensions-builtin\sd_forge_controlnet\install.py:2: DeprecationWarning: pkg_resources is deprecated as an API. See setuptools.pypa.io/en/latest/pkg_resources.html import pkg_resources Launching Web UI with arguments: --directml --skip-torch-cuda-test --gradio-allowed-path '...\StabilityMatrix\Data\Images' Using directml with device: Total VRAM 1024 MB, total RAM 130993 MB pytorch version: 2.4.1+cpu Set vram state to: NORMAL_VRAM Device: privateuseone VAE dtype preferences: [torch.float32] -> torch.float32 CUDA Using Stream: False The cache for model files in Transformers v4.22.0 has been updated. Migrating your old cache. This is a one-time only operation. You can interrupt this and resume the migration later on by calling `transformers.utils.move_cache()`. 0it [00:00, ?it/s] Using sub quadratic optimization for cross attention Using split attention for VAE Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled ControlNet preprocessor location: ...\StabilityMatrix\Data\Packages\stable-diffusion-webui-forge\models\ControlNetPreprocessor Loading additional modules ... done. 2024-11-09 03:31:02,938 - ControlNet - INFO - ControlNet UI callback registered. Model selected: {'checkpoint_info': {'filename': '...\\StabilityMatrix\\Data\\Packages\\stable-diffusion-webui-forge\\models\\Stable-diffusion\\sd\\flux1-dev-bnb-nf4-v2.safetensors', 'hash': 'f0770152'}, 'additional_modules': [], 'unet_storage_dtype': None} Using online LoRAs in FP16: False Running on local URL: 127.0.0.1:7860 To create a public link, set `share=True` in `launch()`. Startup time: 365.2s (prepare environment: 328.7s, launcher: 4.4s, import torch: 15.2s, initialize shared: 0.5s, other imports: 0.7s, list SD models: 0.4s, load scripts: 2.4s, initialize google blockly: 9.5s, create ui: 2.1s, gradio launch: 1.4s). Environment vars changed: {'stream': False, 'inference_memory': 1024.0, 'pin_shared_memory': False} [GPU Setting] You will use 0.00% GPU memory (0.00 MB) to load weights, and use 100.00% GPU memory (1024.00 MB) to do matrix computation. But it does not correctly recognize my VRAM. Both GPUs have 8GB. And It tries to use CUDA.