Got MiniMax H3 video generation running in TensorSharp

A developer has successfully integrated MiniMax H3 video generation into TensorSharp, an inference engine initially designed for local GGUF/LLM inference. This development allows for image-to-video generation locally, using an image as context and a prompt to describe motion. The project aims to converge LLM, multimodal, image, and video inference into a single local engine, reducing the need for separate Python stacks. Future optimizations are focused on VRAM usage, generation speed, quantized-model support, multi-GPU capabilities, and longer video generation.