Speech-to-Maya

Speech-to-Maya is an advanced, local-first voice copilot built for Autodesk Maya, designed to transform how creators build and modify 3D environments through natural language. Traditional 3D pipelines require tedious manual clicking, menu navigation, and repetitive geometry setup. Speech-to-Maya changes the interaction model by answering a core question: “What if you could build a 3D scene by simply telling Maya what you want?” ### Key Features & Architecture - **Real-Time Voice Streaming:** Captures microphone audio and streams it with ultra-low latency using AssemblyAI's WebSocket API for instant transcription. - **Local Privacy & Speed:** Processes commands locally via Ollama running Qwen 2.5, eliminating cloud API bottlenecks and ensuring complete data privacy. - **Procedural Intelligence:** Translates conversational prompts into mathematical loops and procedural layouts (such as multi-step staircases or stacked containers) rather than dropping static assets. - **Live Desktop Bridge:** Dynamically maps human intent into executable `maya.cmds` Python code to modify the active 3D viewport in real-time. - **Conversational Iteration:** Allows artists to build, modify, and refine scenes seamlessly through continuous dialogue rather than isolated single-shot commands. Speech-to-Maya is a local desktop integration tool built for Autodesk Maya and local Ollama instances, fully open-sourced via GitHub.