MultiAct: Text-to-Motion Generation from Composite Text via Tailored Attention Guidance

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sala, Nathan, Abramovich, Ofir, Shamir, Ariel, Cohen-Or, Daniel, Aristidou, Andreas, Raab, Sigal
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916064585056256
author Sala, Nathan
Abramovich, Ofir
Shamir, Ariel
Cohen-Or, Daniel
Aristidou, Andreas
Raab, Sigal
author_facet Sala, Nathan
Abramovich, Ofir
Shamir, Ariel
Cohen-Or, Daniel
Aristidou, Andreas
Raab, Sigal
contents Text-to-motion generation has progressed rapidly in recent years, offering an expressive interface for animation and human-computer interaction. However, current models remain brittle when handling prompts that describe multiple actions occurring at the same time. Rather than realizing all components of a composite description, models frequently prioritize a single dominant action and neglect the rest, leading to incomplete or ambiguous motion. We present MultiAct, an unpaired, inference-time framework for compositional text-to-motion synthesis that operates directly on pretrained motion generators without retraining or architectural modification. Our method counteracts semantic collapse by adaptively amplifying cross-attention scores associated with underrepresented prompt components. We note that effective modulation depends on prompt-specific choices, such as which tokens and layers to target, and introduce a lightweight auxiliary decision scheme that determines the most effective attention-strengthening parametrization. Extensive quantitative and qualitative evaluations demonstrate that MultiAct consistently outperforms existing baselines on composite prompts, achieving improved semantic coverage while preserving motion realism. Project page: https://natsala13.github.io/multiact.github.io.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30925
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MultiAct: Text-to-Motion Generation from Composite Text via Tailored Attention Guidance
Sala, Nathan
Abramovich, Ofir
Shamir, Ariel
Cohen-Or, Daniel
Aristidou, Andreas
Raab, Sigal
Computer Vision and Pattern Recognition
Graphics
Text-to-motion generation has progressed rapidly in recent years, offering an expressive interface for animation and human-computer interaction. However, current models remain brittle when handling prompts that describe multiple actions occurring at the same time. Rather than realizing all components of a composite description, models frequently prioritize a single dominant action and neglect the rest, leading to incomplete or ambiguous motion. We present MultiAct, an unpaired, inference-time framework for compositional text-to-motion synthesis that operates directly on pretrained motion generators without retraining or architectural modification. Our method counteracts semantic collapse by adaptively amplifying cross-attention scores associated with underrepresented prompt components. We note that effective modulation depends on prompt-specific choices, such as which tokens and layers to target, and introduce a lightweight auxiliary decision scheme that determines the most effective attention-strengthening parametrization. Extensive quantitative and qualitative evaluations demonstrate that MultiAct consistently outperforms existing baselines on composite prompts, achieving improved semantic coverage while preserving motion realism. Project page: https://natsala13.github.io/multiact.github.io.
title MultiAct: Text-to-Motion Generation from Composite Text via Tailored Attention Guidance
topic Computer Vision and Pattern Recognition
Graphics
url https://arxiv.org/abs/2605.30925