EDGE: The Diffusion-Based AI Model That Choreographs Realistic Dance Moves & Conveys to Live Dancers

Artificial intelligence is cutting-edge, and dancing is liberating, and a brilliant form of communication and expression. Imagine a world full of possibilities when AI and dance animation come together.

In an exciting update in the Generative AI domain, researchers at Stanford University have developed a new AI model that can create animated human dance moves to go with any song or piece of music. This exciting creation is called Editable Dance Generation or EDGE.

Karen Liu, an Associate Professor Of Computer Science, who led the team toward this exciting development, states that EDGE demonstrates how characters with AI capabilities are capable of adding an additional level of rhythm, melody, and artistry to dance animation that was previously not possible.

Recommended: Meet GPTZero – An App Created by Student that Detects Essays Written by AI 

What is Editable Dance Generation (EDGE) do?

According to the paper, the state-of-the-art dancing generation method Editable Dance Generation (EDGE) creates biologically plausible, authentic dance gestures from input music. In their method, a transformer-based diffusion model is combined with a potent music feature extractor dubbed Jukebox.

Diffusion models can be defined as a type of deep generative model that learns a data distribution by reversing a planned noise-generation process.

The paper explains that in addition to the advantages immediately conferred by the modeling choices, we observe flaws with previous metrics and propose a new metric that captures the physical accuracy of ground contact behaviors without explicit physical modeling.

Karen and her team believe that this diffusion-based model will assist choreographers create dance sequences besides conveying their ideas to live dancers with 2D dance sequences. EDGE can be touted as the new tool that redefines ‘choreographic idea generation and movement planning.’

• Editability is an essential part of the application’s sophisticated features: This tool can create computer-animated dance sequences by enabling animators to intuitively edit any parts of dance moves.
• While the animator can choreograph the character’s particular leg movements, EDGE will automatically align the entire body that is believable, smooth, and also physically plausible for humans.
• The moves will remain consistent according to the animator’s musical choice.
• Users will soon be able to upload their very own tracks and even perform the motions in front of a camera using EDGE.

The editability allows choreographers as well as dancers to constantly enhance their movements by adding specific poses at strategic times. The additional information is subsequently incorporated automatically into the sequence by EDGE.

Recommended: Google’s Bard Now Has Programming Capabilities, Can Assist Users to Code, Debug & Explain 

“We think it’s a really a fun and engaging way for everyone, not just dancers, to express themselves through movement and tap into their own creativity,” Liu adds.

Jonathan Tseng and Rodrigo Castellon, the student collaborators on the project believe that EDGE represents a perfect fusion of technology and movement because of its capacity to produce enthralling dances in response to almost any song. And, it is likely to open up fresh opportunities for physical activity and artistic expression.

EDGE will be formally introduced at the Computer Vision and Pattern Recognition conference in Vancouver, British Columbia, in June.

Meanwhile, if you want to try to explore some fun moves, and watch EDGE generate new dance moves from scratch, go to their website called the EDGE playground.