Welcome to L2MAS: protocol-first Live2D multi-agent animation roadmap

[XucroYuri]

L2MAS is an early open-source prototype for Live2D multi-agent animation generation. The goal is to start with a verifiable local MVP and gradually evolve toward a production-deployable v2.0 architecture.

Current focus

• Keep agents model-agnostic through capability routing.
• Treat cloud and local AI providers as first-class peers.
• Validate the MVP path with deterministic mock providers and local FFmpeg composition.
• Grow toward real provider adapters for local LLMs, ComfyUI workflows, and Live2D runtime rendering.

Good entry points

• Provider adapter: OpenAI-compatible and Ollama local LLM routing #8 Provider adapter: OpenAI-compatible and Ollama local LLM routing
• Provider adapter: ComfyUI character and storyboard visual workflow #9 Provider adapter: ComfyUI character and storyboard visual workflow
• Live2D runtime spike: Cubism render adapter contract #10 Live2D runtime spike: Cubism render adapter contract

Useful links

• README: https://github.com/XucroYuri/L2MAS#readme
• v0.1.0 release: https://github.com/XucroYuri/L2MAS/releases/tag/v0.1.0
• Discovery profile: https://github.com/XucroYuri/L2MAS/blob/main/docs/github/discovery-profile.md
• Two-stage roadmap: https://github.com/XucroYuri/L2MAS/blob/main/docs/architecture/two-stage-roadmap.md

Questions, provider experiments, local deployment notes, and Live2D integration ideas are welcome here.