Type | |
Stats | 160 0 |
Reviews | (24) |
Published | Aug 31, 2024 |
Base Model | |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 ECBA95DB6B |
Protonic+ is a merge of various merged checkpoints that depict scenes of nature and backgrounds (in general) in a very stylistic and pretty way. It's primarily anime-ish with a somewhat 3D style. I merged several checkpoints then merged my merged checkpoints with the same initial merged checkpoints and then with my previously merged checkpoints until I was happy with the result (it's basically a salad of merges).
So basically, this checkpoint is good with scenery & backgrounds, but it is also capable of generating characters, both in SFW and NSFW contexts.
For negative prompting I highly recommend the use of negative embeddings such as FastNegativeV2 or badhandv4 especially if you intend on generating characters.
It's really no different to other SD1.5 checkpoints in terms of what sampling methods and positive prompts to use.
Euler A: 30 steps
DPM++ 2M SDE Karras: >40 steps.
Other samplers: I dunno!
CFG scale: 4-9 for best results
Recommended to use with high res fix enabled and clip skip 2 for optimal results.
How I made my images:
(example)
Prompt: Cyberpunk city, alley, people, neon lights, cinematic
Negative Prompt: FastNegativeV2
Inpaint areas of the image if necessary
Upscale the image using SwinIR_4x by two
Used Windows Photos to remove elements of the image I didn't like and change the color scheme/shadows/contrast/etc.
--> If you'd like to elaborate on your prompts you can try out:
https://www.neuralframes.com/tools/stable-diffusion-prompt-generator
If you like this model or my artwork please consider tipping me :)
https://ko-fi.com/inno_squirrel
Protons are cool.