Should be very feasible. For H3 can use WanGP/MMGP/ComfyUI backend, and for the LLM llama.cpp to maintain state. The idea is mostly about the workflow and straightforward enough that anyone could just ask an agent to build this, but using aforementioned backends, and build an app with a built-in launcher. In fact, someone should try precisely that -- paste this post URL to an agent and ask it to implement!
Should be very feasible. For H3 can use WanGP/MMGP/ComfyUI backend, and for the LLM llama.cpp to maintain state. The idea is mostly about the workflow and straightforward enough that anyone could just ask an agent to build this, but using aforementioned backends, and build an app with a built-in launcher. In fact, someone should try precisely that -- paste this post URL to an agent and ask it to implement!