The simplest way to self-host Photon. Launch a dedicated cloud GPU server running Lab Station OS to download and serve the model using any compatible app or framework.
Direct Download
Download model weights for local inference. Must be used with a compatible app, notebook, or codebase. May run slowly, or not work at all, depending on your system resources, particularly GPU(s) and available VRAM.
Forge is a platform built on top of Stable Diffusion WebUI to make development easier, optimize resource management, speed up inference, and study experimental features.
Train your own LoRAs and finetunes for Stable Diffusion and Flux using this popular GUI for the Kohya trainers.
Model Report
Photographer / Photon
Photon is an image generation model that combines multiple LORA models trained on AI-generated images. It excels at scenes blending mechanical and organic elements, with strong environmental effects. Uses DPM++ 2M Karras sampler, works best with simple prompts at 512x768 resolution.
Explore the Future of AI
Your server, your data, under your control
Photon is a Stable Diffusion checkpoint model focused on generating highly photorealistic images. The model represents a significant advancement in AI image generation through its unique development approach of training and combining multiple LORAs using AI-generated photorealistic images.
Development and Architecture
The model's architecture is built upon careful training of multiple LORA models focused on different aspects of photorealism. These LORAs were then merged with varying weights to create the final checkpoint model. This approach allows Photon to excel at generating highly detailed and realistic images across various subjects.
One notable challenge during development has been hand generation quality, which the creators acknowledge and are actively working to improve in future versions. The current version (v1) demonstrates strong overall photorealistic capabilities while still showing some limitations in this specific area.
Optimal Usage and Parameters
To achieve the best results with Photon, the creators recommend the following parameters:
Simple, straightforward prompts
Specific negative prompts: "cartoon, painting, illustration, (worst quality, low quality, normal quality:2)"
DPM++ 2M Karras sampler
20 steps
CFG scale of 6
Initial generation size of 512x768 or 768x512
For higher resolution outputs, the creators suggest using R-ESRGAN 4x+ for upscaling. It's important to note that negative embeddings should be avoided unless absolutely necessary, as they can potentially impact the model's performance.
Future Development and Licensing
The development roadmap for Photon includes several ambitious goals:
Eliminating the need for negative prompts
Completely resolving hand generation issues
Expanding the training dataset to 5,000-50,000 high-quality AI-generated photorealistic images
The model is released under the CreativeML Open RAIL-M license with a specific addendum. More details about the license can be found in the Photon Model License Addendum.
