A model merge by Inner Reflections that combines two video models whose strengths are complementary. It is the engine of the free AI renderer in free-ai-renderer-comfyui-workflow.
An AI renderer needs one model with three properties:
Every model the author's team tested had only some of these. They were close to giving up before finding a model merge.
The merged model accepts control nets and references together, so they work in tandem. New areas revealed by camera movement still match the reference's style. The author puts it this way: "It's just like the model understands two languages."
Use the merged model as the engine of the AI renderer workflow. Control passes pin the geometry, and the reference image drives the look. See Generative AI rendering with control nets, reference image and prompt and control-pass-choice-sets-model-freedom.