Everything is always so interesting and you were definitely right about not knowing the last style. I'll definitely try it, thanks for this interesting and out of the box suggestion 👌
For videos like this not specific to Fooocus I usually do not since it makes people think it is only specific to Fooocus and hurts it being seen by a larger audience. But I do have my Fooocus playlist which I do add them to. ru-vid.com/group/PLSofK8pvSk5oHH9iNAAemQREADEY7CPp_
Your videos are so good.How do you know all of this stuff? I was wondering, is there somewhere that you can see what the style adds or removes? Also, can you make your own styles to include? And a final random comment; I really wish Fooocus had a tool to easily build Loras.
There is a folder in the Fooocus folder if using Fooocus called sdxl_styles that will have the different json files. You can open those with a text viewer of choice to see the styles. No easy way otherwise. I have a custom GPT that I am working on in ChatGPT that is trained on all that stuff for that reason so I can just ask. I got tired of looking them up that way.
Has anybody ever found a style similar to the "sort of" photorealism of DALL-E/Bing? I use that for the consistency and have a character I've created (she's my profile pic) but when I try to translate her to Fooocus (using face swap or image prompts) they come out "too" real, kind of warts and all. I really want a kind of hyperreal style to keep continuity but the frankly absurd Bing prohibited prompts list is just getting too much. Any suggestions on some good settings or styles that could match that kind of semi realism? Thanks!
@@KLEEBZTECH ooh thank you. I will go look at that. After I watch your video on Loras (I did see one, right? I seem to recall it being on my to watch list.)
Yes I do have one on LoRAs and checkpoints. That may be the solution. Not all models know the same things. Also may find LoRAs for specific hairstyles.
@@KLEEBZTECH thanks. I'm still working my way through your videos (which are a huge help, thank you) and learning a lot about Fooocus through them. Thanks for producing them and for setting me on the right path here. I appreciate it!
Yeah so I discovered that some models it will require you to use the API key which I added a comment to my Colab video about it. You would need to activate one on your account in the settings and add it to the link to download. But my personal way that I like (if you have enough storage space in Google Drive) is to upload to Google Drive and share it so you get the link and use that. By it being in Google Drive it only takes a second for it to grab it. Just discovered that the other day.
Hello, I have been struggling with something for a very long time. Please help. So whenever i create images in fooocus in rundiffussion i get better quality but whenever I do this locally on my pc (4090) i get poor quality results. I have adjusted the same setting , same checkpoint and everything. What am I missing?
I can't think of anything. Have you tried a different install locally? I can't imagine why. Have you tested bringing an image over with metadata from rundiffusion and seeing if you can recreate the same one?
When doing some things like upscaling I have to minimize my browser windows until it gets started or else it takes forever. I am very limited on doing other things on my computer that might use the GPU. Makes it difficult for example when I record for RU-vid when I do anything that requires more than basic image generation. I personally would go higher if getting a new GPU at this point myself but not an option right now for me money wise.