LLM_Plugin

Local Multimodal LLM is an Unreal Engine 5 plugin that enables offline, local inference for multimodal AI within the engine. It provides a suite of capabilities including Large Language Model (LLM) processing, Speech-to-Text (STT), and Text-to-Speech (TTS), along with character memory, routing, and constrained tool calls. Key technical features include: - Native Win64 inference runtimes with support for CPU, Vulkan, and optional CUDA 12 acceleration for NVIDIA GPUs. - Integration with specific models such as Gemma 4 E2B, Parakeet STT, and Pocket TTS. - Blueprint-facing APIs for integration into Unreal Engine projects. - A companion MetaHuman Audio Live Link bridge for synchronizing audio with MetaHuman characters. - Support for provider-neutral voice catalogs and per-character voice selection.