MoLA: Motion Generation and Editing with Latent Diffusion Enhanced by Adversarial Training

TL;DR

MoLA is a novel framework that combines latent diffusion, adversarial training, and a variational autoencoder to enable fast, high-quality, and controllable text-to-motion generation and editing of variable-length motions.

Contribution

It introduces a unified approach for motion generation and editing using latent diffusion enhanced by adversarial training, with a new motion representation for variable-length outputs.

Findings

Adversarial training improves motion generation quality.

The framework supports multiple motion editing tasks.

Motion generation is faster and more controllable.

Abstract

In text-to-motion generation, controllability as well as generation quality and speed has become increasingly critical. The controllability challenges include generating a motion of a length that matches the given textual description and editing the generated motions according to control signals, such as the start-end positions and the pelvis trajectory. In this paper, we propose MoLA, which provides fast, high-quality, variable-length motion generation and can also deal with multiple editing tasks in a single framework. Our approach revisits the motion representation used as inputs and outputs in the model, incorporating an activation variable to enable variable-length motion generation. Additionally, we integrate a variational autoencoder and a latent diffusion model, further enhanced through adversarial training, to achieve high-quality and fast generation. Moreover, we apply a training-free guided generation framework to achieve various editing tasks with motion control inputs. We quantitatively show the effectiveness of adversarial learning in text-to-motion generation, and demonstrate the applicability of our editing framework to multiple editing tasks in the motion domain.