EBE - imagine diffusion
EBE (Enhanced By Ehristoforu) is a modern project whose task is to create the best universal and balanced diffusion models based on SD1.5, SD2.1, SDXL, PonyV6, SD3 and others.
About this project
Who is working on the project
The Project Fluently and DreamDrop team led by ehristoforu are working on this project.
What are the goals of the project
Our project has several key goals:
Creating the most versatile and balanced models
Creating a friendly community in Discord/HuggingFace, where everyone will participate in improving the models
Creation of LoRAs for different tasks
Creating styles for our models
Current roadmap
Our roadmap looks like this:
Creation of EBE based on SD1.5
Creation of EBE based on SD2.1
Creation of EBE based on SDXL
Creation of EBE based on PonyV6
Creation of EBE based on SD3 Medium
Creation of LoRAS for EBE based on SD1.5
Creation of LoRAS for EBE based on SDXL
Creation of LoRAS for EBE based on PonyV6
Creation of LoRAS for EBE based on SD3 Medium
Create a large number of quality styles for EBE models
How our models are created
We use several techniques at once:
Training base models on large datasets
Megring well-selected models
We train LoRAs and integrate them into based models
Is merging harmful to the model?
In our case no, we carefully select and test models for merge, run tests using X/Y/Z plot and much more, so don't worry.
Links
Links to our resources:
Our LoRAs: coming soon
Our styles: coming soon
HuggingFace: here
Our site: coming soon
Our Discord community: coming soon
Thanks 💖
Description
Work done:
The model was trained on a large dataset.
More than 20 best models took part in the merger
Results:
Improved overall aesthetics and quality
Improved understanding and adherence to prompts
Improved tokens: cinematic, depth of field, semi-realistic