t2m4lvo-truebones-zoo

🐉 🐕🐊🦅🐀🦈 T2M4LVO: Text-to-Motion Synthesis for Large-Vocabulary Objects T2M4LVO is a high-quality, human-annotated dataset for text-to-motion synthesis over non-human, diverse skeletal structures, created as part of the ICML 2025 paper How to Move Your Dragon. While limited in scale (approx. 1K motions), it is the first dataset to offer multi-level textual descriptions for a wide variety of animal, creature, and fantasy motions, enabling research beyond human-centric… See the full description on the dataset page:

Source: Hugging Face Hub (1Konny/t2m4lvo-truebones-zoo). Metadata imported from the dataset’s Hub tags.