xGen-VideoSyn-1: High-fidelity Text-to-Video Synthesis with Compressed Representations

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Qin, Can, Xia, Congying, Ramakrishnan, Krithika, Ryoo, Michael, Tu, Lifu, Feng, Yihao, Shu, Manli, Zhou, Honglu, Awadalla, Anas, Wang, Jun, Purushwalkam, Senthil, Xue, Le, Zhou, Yingbo, Wang, Huan, Savarese, Silvio, Niebles, Juan Carlos, Chen, Zeyuan, Xu, Ran, Xiong, Caiming
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909302015393792
author Qin, Can
Xia, Congying
Ramakrishnan, Krithika
Ryoo, Michael
Tu, Lifu
Feng, Yihao
Shu, Manli
Zhou, Honglu
Awadalla, Anas
Wang, Jun
Purushwalkam, Senthil
Xue, Le
Zhou, Yingbo
Wang, Huan
Savarese, Silvio
Niebles, Juan Carlos
Chen, Zeyuan
Xu, Ran
Xiong, Caiming
author_facet Qin, Can
Xia, Congying
Ramakrishnan, Krithika
Ryoo, Michael
Tu, Lifu
Feng, Yihao
Shu, Manli
Zhou, Honglu
Awadalla, Anas
Wang, Jun
Purushwalkam, Senthil
Xue, Le
Zhou, Yingbo
Wang, Huan
Savarese, Silvio
Niebles, Juan Carlos
Chen, Zeyuan
Xu, Ran
Xiong, Caiming
contents We present xGen-VideoSyn-1, a text-to-video (T2V) generation model capable of producing realistic scenes from textual descriptions. Building on recent advancements, such as OpenAI's Sora, we explore the latent diffusion model (LDM) architecture and introduce a video variational autoencoder (VidVAE). VidVAE compresses video data both spatially and temporally, significantly reducing the length of visual tokens and the computational demands associated with generating long-sequence videos. To further address the computational costs, we propose a divide-and-merge strategy that maintains temporal consistency across video segments. Our Diffusion Transformer (DiT) model incorporates spatial and temporal self-attention layers, enabling robust generalization across different timeframes and aspect ratios. We have devised a data processing pipeline from the very beginning and collected over 13M high-quality video-text pairs. The pipeline includes multiple steps such as clipping, text detection, motion estimation, aesthetics scoring, and dense captioning based on our in-house video-LLM model. Training the VidVAE and DiT models required approximately 40 and 642 H100 days, respectively. Our model supports over 14-second 720p video generation in an end-to-end way and demonstrates competitive performance against state-of-the-art T2V models.
format Preprint
id arxiv_https___arxiv_org_abs_2408_12590
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle xGen-VideoSyn-1: High-fidelity Text-to-Video Synthesis with Compressed Representations
Qin, Can
Xia, Congying
Ramakrishnan, Krithika
Ryoo, Michael
Tu, Lifu
Feng, Yihao
Shu, Manli
Zhou, Honglu
Awadalla, Anas
Wang, Jun
Purushwalkam, Senthil
Xue, Le
Zhou, Yingbo
Wang, Huan
Savarese, Silvio
Niebles, Juan Carlos
Chen, Zeyuan
Xu, Ran
Xiong, Caiming
Computer Vision and Pattern Recognition
Artificial Intelligence
We present xGen-VideoSyn-1, a text-to-video (T2V) generation model capable of producing realistic scenes from textual descriptions. Building on recent advancements, such as OpenAI's Sora, we explore the latent diffusion model (LDM) architecture and introduce a video variational autoencoder (VidVAE). VidVAE compresses video data both spatially and temporally, significantly reducing the length of visual tokens and the computational demands associated with generating long-sequence videos. To further address the computational costs, we propose a divide-and-merge strategy that maintains temporal consistency across video segments. Our Diffusion Transformer (DiT) model incorporates spatial and temporal self-attention layers, enabling robust generalization across different timeframes and aspect ratios. We have devised a data processing pipeline from the very beginning and collected over 13M high-quality video-text pairs. The pipeline includes multiple steps such as clipping, text detection, motion estimation, aesthetics scoring, and dense captioning based on our in-house video-LLM model. Training the VidVAE and DiT models required approximately 40 and 642 H100 days, respectively. Our model supports over 14-second 720p video generation in an end-to-end way and demonstrates competitive performance against state-of-the-art T2V models.
title xGen-VideoSyn-1: High-fidelity Text-to-Video Synthesis with Compressed Representations
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2408.12590