Saved in:
Bibliographic Details
Main Authors: Li, Shuowei, Zhao, Yuming, Bhalerao, Parth, Ignat, Oana
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.16716
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916063510265856
author Li, Shuowei
Zhao, Yuming
Bhalerao, Parth
Ignat, Oana
author_facet Li, Shuowei
Zhao, Yuming
Bhalerao, Parth
Ignat, Oana
contents Text-to-video (T2V) generation has rapidly progressed in visual fidelity, yet its ability to faithfully represent multiple cultures within a single prompt remains underexplored. We introduce MAVEN, a multi-agent prompt refinement framework designed to improve cultural fidelity in both mono-cultural and cross-cultural T2V generation. MAVEN decomposes prompts into person, action, and location dimensions, handled by specialized agents operating in parallel or sequentially. To support systematic evaluation, we contribute a new benchmark of 243 culturally grounded prompts and 972 corresponding videos, spanning three cultures (Chinese, American, Romanian), three action categories, and both mono-cultural and cross-cultural scenarios. Evaluations combining CLIP-based metrics, VLM-as-judge assessments, and videoquality measures show that multi-agent refinement, particularly parallel specialization, significantly improves cultural relevance while preserving visual quality and temporal consistency. The dataset and code are available at https://github.com/AIM-SCU/MAVEN
format Preprint
id arxiv_https___arxiv_org_abs_2605_16716
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MAVEN A Multi-Agent Framework for Multicultural Text-to-Video Generation
Li, Shuowei
Zhao, Yuming
Bhalerao, Parth
Ignat, Oana
Computer Vision and Pattern Recognition
Artificial Intelligence
Text-to-video (T2V) generation has rapidly progressed in visual fidelity, yet its ability to faithfully represent multiple cultures within a single prompt remains underexplored. We introduce MAVEN, a multi-agent prompt refinement framework designed to improve cultural fidelity in both mono-cultural and cross-cultural T2V generation. MAVEN decomposes prompts into person, action, and location dimensions, handled by specialized agents operating in parallel or sequentially. To support systematic evaluation, we contribute a new benchmark of 243 culturally grounded prompts and 972 corresponding videos, spanning three cultures (Chinese, American, Romanian), three action categories, and both mono-cultural and cross-cultural scenarios. Evaluations combining CLIP-based metrics, VLM-as-judge assessments, and videoquality measures show that multi-agent refinement, particularly parallel specialization, significantly improves cultural relevance while preserving visual quality and temporal consistency. The dataset and code are available at https://github.com/AIM-SCU/MAVEN
title MAVEN A Multi-Agent Framework for Multicultural Text-to-Video Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.16716