Forum Replies Created
-
Posts
-
It would set you up for success if you find a token that generates objects similar to what you want to train. As always, good training images in a setting similar to what you are going to generate helps.
Welcome!
If you want to be thorough and have a solid foundation, you can go through the Stable Diffusion levels 1 to 4 first. Then you can tackle the comfyui courses.
If you want to get there fast, you can take the comfyui course directly. I don’t repeat the content in the SD courses but link to relevant articles.
The workflows are not ordered in the skill level. You would need to judge which ones you want to study at your level.
The comyfui workflow on the resources page is a good resources to study a particular module in isolation. The comfyui workflows out there usually have massive number of nodes, making them difficult to study and reuse.
Sorry for the late response. The message went into the moderation mailbox.
For comfyui, it is common to see the installation outdated and causing problems. You should update comfyui itself and all nodes by following this:
https://stable-diffusion-art.com/animate-partial-image-comfyui/#Step_0_Load_the_ComfyUI_workflow
If it still doesn’t work, the workflow may need to be revised. Let me know, and I will look into it.
Regarding things not working when changing the prompt, it is common for Stable Diffusion workflows to need some tweaking when changing the prompt. Not all prompts work with a particular workflow. Just like creating art with other mediums, it needs experimentation. The workflow is an example of what is possible. It serves as a good starting point but not a universal solution.
Installing the comfy portable in Windows is the easiest way. You download, unzip and run. You can try that route if you haven’t already.
Comfyui is a bit harder to learn but it also has many pre-made video workflow available.
Hi, have you installed the clip vision checkpoint file and put it in the model folder? Are you able to select it?
Hi, you can use the “Install models from URL” section in the notebook.
Put in Model_from_URL:
Before starting the notebook.
I will add the model to the list later.
Great suggestion! Played with it when it came out but it was too buggy. It should be a good time to revisit now.