MCP is not just expensive to run repeated operations in Maya, it's also too slow. Every command takes time to go back and forth, and that's a problem when you want to work fast during rigging and skinning.
So here's the idea: instead of using MCP for small, repeated actions, why not use a local voice recognition model connected directly to Maya commands? You say the word, and the action happens right away. No waiting, no digging through menus, no setting up hotkeys.
What this fixes for me
- The magic number — I have a magic number for smoothing weights in ngSkinTools: intensity 4.5, iterations 100, given to me by my old supervisor. Instead of typing these numbers every time, I just say "smooth" and it runs with those exact values already set.
- The missing hotkeys — ngSkinTools doesn't have hotkeys for some actions I use a lot, like randomizing influence colors or switching brush mode. Instead of trying to force Maya's hotkey editor to work for these, I just say the command out loud.
This is still early. It uses a local, offline speech model (no internet, no API needed) that matches your voice to a small list of commands, then sends it straight to Maya. I'll keep adding more commands over time.
Code is on GitHub — pecktd/vox_maya.