sharlotte
Member
- Jan 10, 2019
- 323
- 1,749
Best place to take your first steps is at Civitai -Hi guys. I'm just starting to use SD. Can you please tell me what SD-models /any other programms are needed to achieve such photorealism?
Or maybe the selfie picture was made in Midjourney?
Thanks for any help! View attachment 2730364 View attachment 2730364
I recommend you PhotonHi guys. I'm just starting to use SD. Can you please tell me what SD-models /any other programms are needed to achieve such photorealism?
Or maybe the selfie picture was made in Midjourney?
Thanks for any help! View attachment 2730364 View attachment 2730364
That's super nice! I tried to play around it using realistic lineart controlnet and it works better than canny if that helps!Pretty new with using Ai (been using it couple of days) I've been seeing it everywhere though, so thought might as well check it out, for the image below I used one of my old Daz3d renders (that I was never really happy with) plugged into control net set to canny, the output from that was pretty awful (as to be expected) I then sent the awful image to imgtoimg turned up the samples and the resolution and generated a batch of images, the result was much better (still not perfect) I then upscaled and this was the final result, just did a bit of postwork after, I'm relatively happy with it. (I've also added the original daz render below, although the only thing vaguely similar between the images is the pose )
View attachment 2723792 View attachment 2723805
Important note, not all Lora's work well with all checkpoints or other models, this is the fucked up part where you need to find out what combines with what...Best place to take your first steps is at Civitai -You must be registered to see the links. You can see what people are recommending there and in most cases they also give you examples that if you click into them have the prompts used to create the model.
The first thing you want is what we call a model but is called a checkpoint at that site. I use this one for realistic results but other checkpoints are just as if not more popular :-You must be registered to see the links
For instance if you click on the link for the lassie from Game of Thrones you will see the prompt in the bottom right of the screen used to create her. :-You must be registered to see the links
Which then brings you to Loras and Lycos, which are used to shortcut things that you want to insert into your picture. So instead of painfully trying to describe a star-trek uniform in a prompt, you just use a Lora as such :-You must be registered to see the links. You can use loras etc for just about anything, from a celebrity to a specific type of car to a jungle on a different planet
Then you spend the next year down a rabbit hole of things to learn. Like control-net. Or poses. Or TI's Or...or...or...
Thank you very much for the short instruction, it helps me a lot.Best place to take your first steps is at Civitai -You must be registered to see the links. You can see what people are recommending there and in most cases they also give you examples that if you click into them have the prompts used to create the model.
The first thing you want is what we call a model but is called a checkpoint at that site. I use this one for realistic results but other checkpoints are just as if not more popular :-You must be registered to see the links
For instance if you click on the link for the lassie from Game of Thrones you will see the prompt in the bottom right of the screen used to create her. :-You must be registered to see the links
Which then brings you to Loras and Lycos, which are used to shortcut things that you want to insert into your picture. So instead of painfully trying to describe a star-trek uniform in a prompt, you just use a Lora as such :-You must be registered to see the links. You can use loras etc for just about anything, from a celebrity to a specific type of car to a jungle on a different planet
Then you spend the next year down a rabbit hole of things to learn. Like control-net. Or poses. Or TI's Or...or...or...
Thank you, I will definitely try this model!I recommend you PhotonYou must be registered to see the linksas a starting point.
A LoRA is an extra bit of info loaded separately into the Checkpoint, assuming you're using Automatic 1111, you load the Checkpoint from the dropdown at the top left and you select the LoRA either by putting the LoRA name and weight in the prompt as <lora-name of lora:weight> or by using the Additional Networks extension (I don't recommend the latter for a beginner) and applying them there.Thank you very much for the short instruction, it helps me a lot.
Did I understand correctly that it does not matter in what order to use "Checkpoint" and "Lora"?
View attachment 2734527
and other example:
View attachment 2734530
And one more question. Prompts are used only for Checkpoint or do they need to be prescribed for Lora too?
Thank you, I will definitely try this model!
Depends on your hardware.Fellas. How do I learn creating AI porn?
I've got an i7 and 3080tiDepends on your hardware.
For a basic 'Dummies Guide', I'd highly recommend the first post on this thread