Home > Research > Publications & Outputs > DreamFactory

Research

Electronic data

  • 2408.11788v1

    Final published version, 7.03 MB, PDF document

Links

Keywords

View graph of relations

DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework

Research output: Working paper › Preprint

Published

Standard

DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework. / Xie, Zhifei; Tang, Daniel; Tan, Dingwei et al.
Arxiv, 2024.

Research output: Working paper › Preprint

Harvard

Xie, Z, Tang, D, Tan, D, Klein, J, Bissyand, TF & Ezzini, S 2024 'DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework' Arxiv. <https://arxiv.org/abs/2408.11788>

APA

Xie, Z., Tang, D., Tan, D., Klein, J., Bissyand, T. F., & Ezzini, S. (2024). DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework. Arxiv. https://arxiv.org/abs/2408.11788

Vancouver

Xie Z, Tang D, Tan D, Klein J, Bissyand TF, Ezzini S. DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework. Arxiv. 2024 Aug 21.

Author

Xie, Zhifei ; Tang, Daniel ; Tan, Dingwei et al. / DreamFactory : Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework. Arxiv, 2024.

Bibtex

@techreport{529b3e3da3764396b0ec73495650ca98,
title = "DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework",
abstract = "Current video generation models excel at creating short, realistic clips, but struggle with longer, multi-scene videos. We introduce \texttt{DreamFactory}, an LLM-based framework that tackles this challenge. \texttt{DreamFactory} leverages multi-agent collaboration principles and a Key Frames Iteration Design Method to ensure consistency and style across long videos. It utilizes Chain of Thought (COT) to address uncertainties inherent in large language models. \texttt{DreamFactory} generates long, stylistically coherent, and complex videos. Evaluating these long-form videos presents a challenge. We propose novel metrics such as Cross-Scene Face Distance Score and Cross-Scene Style Consistency Score. To further research in this area, we contribute the Multi-Scene Videos Dataset containing over 150 human-rated videos.",
keywords = "cs.AI, cs.CL, cs.CV, cs.SE, TsingHua University",
author = "Zhifei Xie and Daniel Tang and Dingwei Tan and Jacques Klein and Bissyand, {Tegawend F.} and Saad Ezzini",
note = "13 pages, 8 figures",
year = "2024",
month = aug,
day = "21",
language = "English",
publisher = "Arxiv",
type = "WorkingPaper",
institution = "Arxiv",

}

RIS

TY - UNPB

T1 - DreamFactory

T2 - Pioneering Multi-Scene Long Video Generation with a Multi-Agent Framework

AU - Xie, Zhifei

AU - Tang, Daniel

AU - Tan, Dingwei

AU - Klein, Jacques

AU - Bissyand, Tegawend F.

AU - Ezzini, Saad

N1 - 13 pages, 8 figures

PY - 2024/8/21

Y1 - 2024/8/21

N2 - Current video generation models excel at creating short, realistic clips, but struggle with longer, multi-scene videos. We introduce \texttt{DreamFactory}, an LLM-based framework that tackles this challenge. \texttt{DreamFactory} leverages multi-agent collaboration principles and a Key Frames Iteration Design Method to ensure consistency and style across long videos. It utilizes Chain of Thought (COT) to address uncertainties inherent in large language models. \texttt{DreamFactory} generates long, stylistically coherent, and complex videos. Evaluating these long-form videos presents a challenge. We propose novel metrics such as Cross-Scene Face Distance Score and Cross-Scene Style Consistency Score. To further research in this area, we contribute the Multi-Scene Videos Dataset containing over 150 human-rated videos.

AB - Current video generation models excel at creating short, realistic clips, but struggle with longer, multi-scene videos. We introduce \texttt{DreamFactory}, an LLM-based framework that tackles this challenge. \texttt{DreamFactory} leverages multi-agent collaboration principles and a Key Frames Iteration Design Method to ensure consistency and style across long videos. It utilizes Chain of Thought (COT) to address uncertainties inherent in large language models. \texttt{DreamFactory} generates long, stylistically coherent, and complex videos. Evaluating these long-form videos presents a challenge. We propose novel metrics such as Cross-Scene Face Distance Score and Cross-Scene Style Consistency Score. To further research in this area, we contribute the Multi-Scene Videos Dataset containing over 150 human-rated videos.

KW - cs.AI

KW - cs.CL

KW - cs.CV

KW - cs.SE

KW - TsingHua University

M3 - Preprint

BT - DreamFactory

PB - Arxiv

ER -