FFUSION AI - 768 BaSE Public alpha Release

Model Overview: Unleashing the Power of Imagination!

FFUSION AI is a state-of-the-art image generation and transformation tool, developed around the leading Latent Diffusion Model. Leveraging Stable Diffusion 2.1, FFUSION AI converts your prompts into captivating artworks. Discover an imaginative landscape where ideas come to life in vibrant, surreal visuals.

Developed by: Idle Stoev, Source Code Bulgaria, Praesidium CX & BlackSwan Technologies
Shared by: FFusion AI
Model type: Diffusion-based text-to-image generation model
Language(s) (NLP): English
License: CreativeML Open RAIL++-M License

Model Use: Enabling Creativity and Exploring AI Frontiers

Designed for research and artistic exploration, FFUSION AI serves as a versatile tool in a variety of scenarios:

Investigating biases and constraints in generative models
Unleashing creativity in artistic endeavors
Infusing AI-driven innovation into educational or creative tools
Furthering research in the exciting field of generative models
Repository: https://github.com/1e-2
Demo: https://huggingface.co/spaces/FFusion/FFusionAI-Streamlit-Playground

Out-of-Scope Use and Prohibited Misuse:

Generating factually inaccurate representations of people or events
Inflicting harm or spreading malicious content such as demeaning, dehumanizing, or offensive imagery
Creating harmful stereotypes or spreading discrimination
Impersonating individuals without their consent
Disseminating non-consensual explicit content or misinformation
Violating copyrights or usage terms of licensed material

Model Limitations and Bias

While our model brings us closer to the future of AI-driven creativity, there are several limitations:

Achieving perfect photorealism or surrealism is still an ongoing challenge.
Rendering legible text could be difficult without further ~30min training on your brand.
Accurate generation of human faces, especially far away faces, is not guaranteed (yet).

Model Releases

We are thrilled to announce:

Version 512 Beta: Featuring LiTE and MiD BFG model variations
Version 768 Alpha: BaSE, FUSION, and FFUSION models with enhanced training capabilities, including LoRa, LyCORIS, Dylora & Kohya-ss/sd-scripts.
Version 768 BaSE: A BaSE Ready model for easy applying more than 200 build op LoRA models trained along the way.

Environmental Impact

In line with our commitment to sustainability, FFUSION AI has been designed with carbon efficiency in mind:

Hardware Type: A100 PCIe 40GB
Hours used: 1190
Cloud Provider: CoreWeave & Runpod (official partner)
Compute Region: US Cyxtera Chicago Data Center - ORD1 / EU - CZ & EU - RO
Carbon Emitted: 124.95 kg of CO2 (calculated via Machine Learning Impact calculator)

That said all LoRA and further models are based on initial training.