ChoreoAI

ChoreoAI is a multimodal-to-motion pipeline for generating 3D dance pose sequences from text prompts. It includes model inference, dataset lifecycle tooling, and a lightweight HTTP API.

Project layout

src/choreoai/
  api/        # HTTP transport layer (FastAPI app and schemas)
  config/     # environment-backed application settings
  core/       # domain entities and validation rules
  services/   # orchestration layer used by CLI/API
  encoders/   # representation learning models
  generator/  # diffusion model components
  utils/      # shared cross-cutting helpers

Quick start

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pip install -e .

Run CLI:

choreoai --help
choreoai summarize --root data/dataset

Run API:

uvicorn choreoai.api:app --reload

Run tests:

pytest

Key decisions

• Keep model code mostly unchanged and refactor around it via services.
• Preserve existing command behavior while reducing adapter-level duplication.
• Add practical validation and telemetry hooks without over-abstracting.

Notes

• .env.example documents supported runtime environment variables.
• Dataset preprocessing still favors readability over maximum throughput.