Embody

Based on the available evidence, Embody is best understood as an open-source bridge that exposes TouchDesigner to AI workflows, not as a polished end-to-end AI visual generation product. The GitHub repo explicitly says it can build, wire, and debug networks with AI, plus provide git-diffable externalization. That supports the positioning, but not yet strong claims about reliability on complex production work.

In practical terms, it appears to let an LLM interact with TouchDesigner through an MCP server so the model can help assemble node graphs, connect operators, troubleshoot issues, and externalize parts of a project into forms that are easier to inspect in Git. A better analogy is not “text-to-video software,” but “an AI-operable development interface for TouchDesigner.” That suggests workflow augmentation rather than automatic final-content generation.

On barriers and cost, the current evidence is mostly the official GitHub repository description.