CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects

Customized text-to-video generation aims to generate high-quality videos guided by text prompts and subject references. Current approaches for personalizing text-to-video generation suffer from tackling multiple subjects, which is a more challenging and practical scenario. In this work, our aim is t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Zhao, Li, Aoxue, Zhu, Lingting, Guo, Yong, Dou, Qi, Li, Zhenguo
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Wang, Zhao
Li, Aoxue
Zhu, Lingting
Guo, Yong
Dou, Qi
Li, Zhenguo
description Customized text-to-video generation aims to generate high-quality videos guided by text prompts and subject references. Current approaches for personalizing text-to-video generation suffer from tackling multiple subjects, which is a more challenging and practical scenario. In this work, our aim is to promote multi-subject guided text-to-video customization. We propose CustomVideo, a novel framework that can generate identity-preserving videos with the guidance of multiple subjects. To be specific, firstly, we encourage the co-occurrence of multiple subjects via composing them in a single image. Further, upon a basic text-to-video diffusion model, we design a simple yet effective attention control strategy to disentangle different subjects in the latent space of diffusion model. Moreover, to help the model focus on the specific area of the object, we segment the object from given reference images and provide a corresponding object mask for attention learning. Also, we collect a multi-subject text-to-video generation dataset as a comprehensive benchmark, with 63 individual subjects from 13 different categories and 68 meaningful pairs. Extensive qualitative, quantitative, and user study results demonstrate the superiority of our method compared to previous state-of-the-art approaches. The project page is https://kyfafyd.wang/projects/customvideo.
doi_str_mv 10.48550/arxiv.2401.09962
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2401_09962</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2401_09962</sourcerecordid><originalsourceid>FETCH-LOGICAL-a672-f869807f11096e4f929e5bdd6a69a1fe1e8cd538f74a3112a74670c7e964ae2a3</originalsourceid><addsrcrecordid>eNotz01OwzAUBGBvWFSFA3RVX8DBdhz_sEMRtEitWBB1G70mz61RmlSJA4XTI9KuRqORRvoIWQieKJtl_BH6S_hKpOIi4c5pOSPrfBxid9qFGrsnei3hN7QHWuAlstixaaIrbLGHGLqWfod4pNuxieHcIP0Y959YxeGe3HloBny45ZwUry9Fvmab99Vb_rxhoI1k3mpnufFCcKdReScdZvu61qAdCI8CbVVnqfVGQSqEBKO04ZVBpxWghHROltfbiVKe-3CC_qf8J5UTKf0D2NJG9A</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects</title><source>arXiv.org</source><creator>Wang, Zhao ; Li, Aoxue ; Zhu, Lingting ; Guo, Yong ; Dou, Qi ; Li, Zhenguo</creator><creatorcontrib>Wang, Zhao ; Li, Aoxue ; Zhu, Lingting ; Guo, Yong ; Dou, Qi ; Li, Zhenguo</creatorcontrib><description>Customized text-to-video generation aims to generate high-quality videos guided by text prompts and subject references. Current approaches for personalizing text-to-video generation suffer from tackling multiple subjects, which is a more challenging and practical scenario. In this work, our aim is to promote multi-subject guided text-to-video customization. We propose CustomVideo, a novel framework that can generate identity-preserving videos with the guidance of multiple subjects. To be specific, firstly, we encourage the co-occurrence of multiple subjects via composing them in a single image. Further, upon a basic text-to-video diffusion model, we design a simple yet effective attention control strategy to disentangle different subjects in the latent space of diffusion model. Moreover, to help the model focus on the specific area of the object, we segment the object from given reference images and provide a corresponding object mask for attention learning. Also, we collect a multi-subject text-to-video generation dataset as a comprehensive benchmark, with 63 individual subjects from 13 different categories and 68 meaningful pairs. Extensive qualitative, quantitative, and user study results demonstrate the superiority of our method compared to previous state-of-the-art approaches. The project page is https://kyfafyd.wang/projects/customvideo.</description><identifier>DOI: 10.48550/arxiv.2401.09962</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2024-01</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,782,887</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2401.09962$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2401.09962$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Wang, Zhao</creatorcontrib><creatorcontrib>Li, Aoxue</creatorcontrib><creatorcontrib>Zhu, Lingting</creatorcontrib><creatorcontrib>Guo, Yong</creatorcontrib><creatorcontrib>Dou, Qi</creatorcontrib><creatorcontrib>Li, Zhenguo</creatorcontrib><title>CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects</title><description>Customized text-to-video generation aims to generate high-quality videos guided by text prompts and subject references. Current approaches for personalizing text-to-video generation suffer from tackling multiple subjects, which is a more challenging and practical scenario. In this work, our aim is to promote multi-subject guided text-to-video customization. We propose CustomVideo, a novel framework that can generate identity-preserving videos with the guidance of multiple subjects. To be specific, firstly, we encourage the co-occurrence of multiple subjects via composing them in a single image. Further, upon a basic text-to-video diffusion model, we design a simple yet effective attention control strategy to disentangle different subjects in the latent space of diffusion model. Moreover, to help the model focus on the specific area of the object, we segment the object from given reference images and provide a corresponding object mask for attention learning. Also, we collect a multi-subject text-to-video generation dataset as a comprehensive benchmark, with 63 individual subjects from 13 different categories and 68 meaningful pairs. Extensive qualitative, quantitative, and user study results demonstrate the superiority of our method compared to previous state-of-the-art approaches. The project page is https://kyfafyd.wang/projects/customvideo.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz01OwzAUBGBvWFSFA3RVX8DBdhz_sEMRtEitWBB1G70mz61RmlSJA4XTI9KuRqORRvoIWQieKJtl_BH6S_hKpOIi4c5pOSPrfBxid9qFGrsnei3hN7QHWuAlstixaaIrbLGHGLqWfod4pNuxieHcIP0Y959YxeGe3HloBny45ZwUry9Fvmab99Vb_rxhoI1k3mpnufFCcKdReScdZvu61qAdCI8CbVVnqfVGQSqEBKO04ZVBpxWghHROltfbiVKe-3CC_qf8J5UTKf0D2NJG9A</recordid><startdate>20240118</startdate><enddate>20240118</enddate><creator>Wang, Zhao</creator><creator>Li, Aoxue</creator><creator>Zhu, Lingting</creator><creator>Guo, Yong</creator><creator>Dou, Qi</creator><creator>Li, Zhenguo</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240118</creationdate><title>CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects</title><author>Wang, Zhao ; Li, Aoxue ; Zhu, Lingting ; Guo, Yong ; Dou, Qi ; Li, Zhenguo</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a672-f869807f11096e4f929e5bdd6a69a1fe1e8cd538f74a3112a74670c7e964ae2a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Wang, Zhao</creatorcontrib><creatorcontrib>Li, Aoxue</creatorcontrib><creatorcontrib>Zhu, Lingting</creatorcontrib><creatorcontrib>Guo, Yong</creatorcontrib><creatorcontrib>Dou, Qi</creatorcontrib><creatorcontrib>Li, Zhenguo</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Wang, Zhao</au><au>Li, Aoxue</au><au>Zhu, Lingting</au><au>Guo, Yong</au><au>Dou, Qi</au><au>Li, Zhenguo</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects</atitle><date>2024-01-18</date><risdate>2024</risdate><abstract>Customized text-to-video generation aims to generate high-quality videos guided by text prompts and subject references. Current approaches for personalizing text-to-video generation suffer from tackling multiple subjects, which is a more challenging and practical scenario. In this work, our aim is to promote multi-subject guided text-to-video customization. We propose CustomVideo, a novel framework that can generate identity-preserving videos with the guidance of multiple subjects. To be specific, firstly, we encourage the co-occurrence of multiple subjects via composing them in a single image. Further, upon a basic text-to-video diffusion model, we design a simple yet effective attention control strategy to disentangle different subjects in the latent space of diffusion model. Moreover, to help the model focus on the specific area of the object, we segment the object from given reference images and provide a corresponding object mask for attention learning. Also, we collect a multi-subject text-to-video generation dataset as a comprehensive benchmark, with 63 individual subjects from 13 different categories and 68 meaningful pairs. Extensive qualitative, quantitative, and user study results demonstrate the superiority of our method compared to previous state-of-the-art approaches. The project page is https://kyfafyd.wang/projects/customvideo.</abstract><doi>10.48550/arxiv.2401.09962</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2401.09962
ispartof
issn
language eng
recordid cdi_arxiv_primary_2401_09962
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
title CustomVideo: Customizing Text-to-Video Generation with Multiple Subjects
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-04T17%3A51%3A48IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=CustomVideo:%20Customizing%20Text-to-Video%20Generation%20with%20Multiple%20Subjects&rft.au=Wang,%20Zhao&rft.date=2024-01-18&rft_id=info:doi/10.48550/arxiv.2401.09962&rft_dat=%3Carxiv_GOX%3E2401_09962%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true