|
|
--- |
|
|
base_model: |
|
|
- lodestones/Chroma1-HD |
|
|
- lodestones/Chroma1-Base |
|
|
tags: |
|
|
- imagegen |
|
|
- chroma |
|
|
pipeline_tag: text-to-image |
|
|
--- |
|
|
This is a merge of Chroma HD + Chroma Base + Accelerators + Flan T5 XXL + FluxVAE to give you everything you need to generate images quickly and easily. |
|
|
|
|
|
I highly recommend adding "High quality, sharp details" at the end of prompts (or "High quality anime" for anime style). |
|
|
|
|
|
Use a CFG of 1, 8 steps for the fastest results at good quality. |
|
|
|
|
|
**Are you getting anime when you don't want it?** Add "Kodak Portra 400 35mm digital camera photo." to your prompt. If it still needs help, try adding "... camera photo from Flikr." |
|
|
|
|
|
**V3:** Tweaked the acclerators and added more "Chroma-HD" into the mix. This one seems to play nicer with sa_solver/beta. |
|
|
|
|
|
**V4:** Added realistic dataset LORA (v4) + 2k_test LORA, then tweaked accelerators a bit more. er_sde/beta recommended. |
|
|
|
|
|
 |
|
|
|
|
|
FP8 precision. |
