I already run ComfyUI on RunPod and have a complex, mostly working graph, but I now need to fold in a brand-new module that lets me adjust facial expressions. The job is purely about integrating this new capability into the existing workflow, not rebuilding everything from scratch. What I want inside the graph is clear: • Real-time expression changes • Batch processing of images • Custom expression templates You’ll be working inside the RunPod environment, so part of the task is making sure every node, custom extension and model path resolves cleanly there. If a bespoke node or Python script is required to drive the expression edits, we will wire it in together and confirm it survives container restarts. Deliverables • Updated ComfyUI workflow file (and any custom nodes/scripts) ready to import • Step-by-step notes on preparing the RunPod container, installing dependencies and triggering the new features • A quick screen-share or recorded walkthrough proving real-time and batch runs both work with the supplied test images Sometimes the workflow gives slightly odd output, so the freelancer should add a node at the end which could have Gemini(or any other method if possible) to interrogate the quality of the output. If the output is good then it can save the image but if it’s not, it should be regenerated. If you thrive on intricate ComfyUI graphs and know your way around RunPod GPUs, I’d love your help ironing this in.