A collection of custom pipelines and developer scripts designed to streamline creative rendering workflows. By connecting Local ComfyUI nodes, Stable Diffusion APIs, and Gemini LLMs, this toolset automates prompt expansion, model swapping, and post-processing filters.
Leverages Gemini models to expand basic 5-word artist prompts into rich, stylistically coherent descriptions, optimized for Stable Diffusion XL (SDXL) text encoders.
Allows queueing a single expanded prompt against multiple Checkpoint models (Anime, Realistic, Cinematic) and LoRAs sequentially to compare style variations automatically.
Applies custom edge-detection and luminance filters in Python to isolate subjects, preparing render canvases for latent space details during secondary high-res passes.
A headless Python SDK that loads JSON workspace configurations directly, pushes prompts to the queue via WebSockets, and downloads final output frames locally.
The core pipeline uses asynchronous queue mechanics in Python to prevent GPU thread locking, feeding image inputs through ControlNet pipelines sequentially.
Packaged in a multi-stage Docker environment compiled with CUDA toolkit dependencies, making it simple to deploy on local workstation hardware or cloud GPU instances.
Features webhook support to trigger automated batch generations directly from chat interfaces (e.g. Discord bot commands or custom web panels).