A new way of creating 3D human movements from text. Use of hierarchical quantization and bidirectional transformers
