Overview

Everything at a glance

Render farm

Delivery queue

Latest renders

People

Real-person LoRAs — 15+ varied photos each

Elf lineup

One LoRA per elf · parents pick at signup · animate any of them with the same video pipeline

Training configs

~/flux/*.yaml — ai-toolkit LoRA jobs on the 3080
ConfigTriggerAction

Live jobs


        
LoRA library
Finished weights
FileSourceSize

      

    


    
    

      
Studio outputs
Samples & renders from ~/flux/output — newest first

        

    


    
    

      
Customers
Elf-service families — elf pick, plan, notes

        
FamilyElfPersonalityPlanStatusActions

    


    
    

      
Delivery queue
Parent update → script → postcard / video telegram

        
CustomerKindUpdateStatusActions

    


    
    

      

        
GPUs

        
Box

      

      
Pipeline
How a delivery flows through the farm

        

          Parent update → script (local LLM) → Flux still on 3080 (elf LoRA) → voice on 3080 (cloned elf voice) → animate on big-VRAM card → lip-sync → mux → send. Nothing leaves the house.