FLAME
Free-form language-based human motion synthesis and editing with diffusion models.
During my KakaoBrain research internship, I contributed across FLAME’s full research lifecycle and published the work as first author at AAAI 2023, where it received an oral presentation. FLAME is a diffusion-based model that generates human motion from free-form language and edits selected frames or joints without task-specific fine-tuning. Its transformer architecture supports variable-length motion and language conditioning for synthesis, editing, prediction, and in-betweening. (Kim et al., 2023)
