So everyone is abuzz with the changes to Buzz, and the overall operation and future of CivitAI. I decided to throw in my 2 cents, as maybe someone will find my perspective helpful.
First, not ONE single picture that I have posted was generated using CivitAI. As I write this article, that is 8702 images. I have ALWAYS generated "off grid", as it were, with no restrictions save what my imagination and the matrix can produce together. YOU CAN TOO!
If you are confused and would like some guidance on how to get started, this is the first place I went, and I suggest you do as well: https://www.reddit.com/r/StableDiffusion/wiki/index/
There is a lot of information there, and it can be hard to take in. However, you can make it happen with a some time and effort. There are plenty of pages dedicated to certain topics, such as renting processor time on 3rd party websites, that can be really helpful.
The next step I would recommend is this: https://lykos.ai/
The program, "Stability Matrix" is an all-in-one setup to get the different checkpoints up and running on your own system. It is very easy to learn, and I would highly recommend it. I personally use Forge, but ComfyUI is the most popular, just harder to learn.
OK, this article is not going to be a technical manual on how to install everything. It doesn't need to be, because Stability Matrix will handle all that for you. However, I will list computer specs you will need to get moving:
1) You will need a Solid State Drive, and you will need somewhere around 300 gigabytes of memory free on it. That is a conservative estimate, I am currently using 475 gigs. But that includes unsorted pictures, unused checkpoints I really don't need, and forgotten style Loras that won't work well but sounded neat.
2) A newer Nivida based graphics card. I am running a 4060 Ti, which has 16 gigs of Vram on it. The more Vram you can afford, the better.
3) Two monitors can be really helpful, but not strictly necessary.
4) Time and patients to learn everything. I swear it isn't has hard as it seems, and if you understand prompting and generating from working on CivitAI, you will have a leg up. But it needs to be stressed, that practice makes perfect, and you will need practice.
A few specifics:
SD 1.5 checkpoints need 6-8 gigs Video Ram, and 16 gigs (regular) Ram. Most people don't use 1.5 anymore, but there it is
SDXL (Which is Pony & illustrious/Noob) Needs 12-16 Gigs Vram and 32 gigs Ram
Flux/Chroma needs 24 gigs Vram, and 64 gigs regular ram
The newer video generators (like Wan 2.2/3) are using the same architecture as Flux, and will need the same specs
If your computer doesn't meet the Ram requirements, it will STILL WORK. But, it will be much slower. The slower it is to generate, the harder it will be to practice, which will make everything less fun.
It takes me about 90 seconds to generate one single image, just like I post. That includes the 50 base steps, the 35 Hires steps, and a Adetailer Face and a Adetailer Penis.
If I had 8 gigs Vram, the time would be at least 200 seconds per picture. It would function the same, everything would work just right. But the system would need to constantly move information around between the SSD and Ram/Vram, since ram couldn't handle the entire matrix on its own. That adds the time.
As I said, this article is to get you thinking about ways around CivitAI, and not a technical manual. If anyone has any questions, you are free to message me here, or if you prefer [email protected].
If there is a lot of interest, I may do another "getting started with Forge" article, but I suspect the information out there is plenty good if you choose to take a look.
By the way, I am going to keep posting 3-4 days a week as I have been. I personally won't see any functionality difference between today and last week, so I have no reason to quit the site. If it becomes a ghost town, yeah maybe I will stop bothering.
Good luck, and have fun!