Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Lumiere: A Space-Time Diffusion Model for Video Generation (2024)
Latte: Latent Diffusion Transformer for Video Generation (2024)
ConsistI2V: Enhancing Visual Consistency for Image-to-Video Generation (2024)
UniVG: Towards UNIfied-modal Video Generation (2024)
Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization (2024)

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend

\n","updatedAt":"2024-02-24T01:21:12.336Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":264}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7035295367240906},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2402.14797","authors":[{"_id":"65d80312685624d5f2dfed19","name":"Willi Menapace","hidden":false},{"_id":"65d80312685624d5f2dfed1a","user":{"_id":"64276311eb9a0ed86180715b","avatarUrl":"/avatars/76f933cd549f10e5e2db379de235d304.svg","isPro":false,"fullname":"Aliaksandr Siarohin","user":"aliaksandr-siarohin","type":"user"},"name":"Aliaksandr Siarohin","status":"admin_assigned","statusLastChangedAt":"2024-02-23T12:31:47.387Z","hidden":false},{"_id":"65d80312685624d5f2dfed1b","user":{"_id":"63610db13c7147fae7de88e3","avatarUrl":"/avatars/d7e97a16cfee39e1e50d7a5b747876f1.svg","isPro":false,"fullname":"Ivan Skorokhodov","user":"universome","type":"user"},"name":"Ivan Skorokhodov","status":"admin_assigned","statusLastChangedAt":"2024-02-23T12:31:54.000Z","hidden":false},{"_id":"65d80312685624d5f2dfed1c","user":{"_id":"635764d4d11d85992d00e014","avatarUrl":"/avatars/f4234ec5c56ac956545643751c5db9aa.svg","isPro":false,"fullname":"Ekaterina Deyneka","user":"edeyneka","type":"user"},"name":"Ekaterina Deyneka","status":"admin_assigned","statusLastChangedAt":"2024-02-23T12:32:00.729Z","hidden":false},{"_id":"65d80312685624d5f2dfed1d","user":{"_id":"6330f570b68c7453d2ee8945","avatarUrl":"/avatars/5b394a72f277aa6c886b21b477010ab0.svg","isPro":false,"fullname":"Tsai-Shien Chen","user":"tschen","type":"user"},"name":"Tsai-Shien Chen","status":"admin_assigned","statusLastChangedAt":"2024-02-23T12:32:06.809Z","hidden":false},{"_id":"65d80312685624d5f2dfed1e","name":"Anil Kag","hidden":false},{"_id":"65d80312685624d5f2dfed1f","name":"Yuwei Fang","hidden":false},{"_id":"65d80312685624d5f2dfed20","name":"Aleksei Stoliar","hidden":false},{"_id":"65d80312685624d5f2dfed21","name":"Elisa Ricci","hidden":false},{"_id":"65d80312685624d5f2dfed22","user":{"_id":"61f19829233c91cbd2f79e70","avatarUrl":"/avatars/a0735a94542b4f7cda5aed8bc4be0538.svg","isPro":false,"fullname":"Jian Ren","user":"alanspike","type":"user"},"name":"Jian Ren","status":"claimed_verified","statusLastChangedAt":"2024-02-24T14:37:00.550Z","hidden":false},{"_id":"65d80312685624d5f2dfed23","name":"Sergey Tulyakov","hidden":false}],"publishedAt":"2024-02-22T18:55:08.000Z","submittedOnDailyAt":"2024-02-22T23:59:44.106Z","title":"Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video\n Synthesis","submittedOnDailyBy":{"_id":"60f1abe7544c2adfd699860c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg","isPro":false,"fullname":"AK","user":"akhaliq","type":"user"},"summary":"Contemporary models for generating images show remarkable quality and\nversatility. Swayed by these advantages, the research community repurposes them\nto generate videos. Since video content is highly redundant, we argue that\nnaively bringing advances of image models to the video generation domain\nreduces motion fidelity, visual quality and impairs scalability. In this work,\nwe build Snap Video, a video-first model that systematically addresses these\nchallenges. To do that, we first extend the EDM framework to take into account\nspatially and temporally redundant pixels and naturally support video\ngeneration. Second, we show that a U-Net - a workhorse behind image generation\n- scales poorly when generating videos, requiring significant computational\noverhead. Hence, we propose a new transformer-based architecture that trains\n3.31 times faster than U-Nets (and is ~4.5 faster at inference). This allows us\nto efficiently train a text-to-video model with billions of parameters for the\nfirst time, reach state-of-the-art results on a number of benchmarks, and\ngenerate videos with substantially higher quality, temporal consistency, and\nmotion complexity. The user studies showed that our model was favored by a\nlarge margin over the most recent methods. See our website at\nhttps://snap-research.github.io/snapvideo/.","upvotes":21,"discussionId":"65d80318685624d5f2dfee4f","ai_summary":"Snap Video, a transformer-based model, addresses video generation challenges by extending EDM for spatial and temporal redundancy, achieving superior quality, consistency, and speed compared to existing U-Net based methods.","ai_keywords":["EDM framework","U-Net","transformer-based architecture","text-to-video model","motion fidelity","visual quality","temporal consistency","motion complexity"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"635964636a61954080850e1d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/635964636a61954080850e1d/0bfExuDTrHTtm8c-40cDM.png","isPro":false,"fullname":"William Lamkin","user":"phanes","type":"user"},{"_id":"652b83b73b5997ed71a310f2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/652b83b73b5997ed71a310f2/ipCpdeHUp4-0OmRz5z8IW.png","isPro":false,"fullname":"Rui Zhao","user":"ruizhaocv","type":"user"},{"_id":"62e54f0eae9d3f10acb95cb9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e54f0eae9d3f10acb95cb9/VAyk05hqB3OZWXEZW-B0q.png","isPro":true,"fullname":"mrfakename","user":"mrfakename","type":"user"},{"_id":"6538119803519fddb4a17e10","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6538119803519fddb4a17e10/ffJMkdx-rM7VvLTCM6ri_.jpeg","isPro":false,"fullname":"samusenps","user":"samusenps","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"6431ae89dec2a70d8132faa4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6431ae89dec2a70d8132faa4/GdXPRTzHiAlSqXodnmbIs.png","isPro":false,"fullname":"C0nsumption","user":"C0nsumption","type":"user"},{"_id":"61e7c06064d3c6c929057bee","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61e7c06064d3c6c929057bee/QxULx1EA1bgmjXxupQX4B.jpeg","isPro":false,"fullname":"蓋瑞王","user":"gary109","type":"user"},{"_id":"61f44bab7eba274ea80b74ce","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61f44bab7eba274ea80b74ce/BRbKX1jephdZ7D44FATl4.jpeg","isPro":false,"fullname":"Hyoung-Kyu Song","user":"deepkyu","type":"user"},{"_id":"648eb1eb59c4e5c87dc116e0","avatarUrl":"/avatars/c636cea39c2c0937f01398c94ead5dad.svg","isPro":false,"fullname":"fdsqefsgergd","user":"T-representer","type":"user"},{"_id":"63d4c8ce13ae45b780792f32","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1675778487155-63d4c8ce13ae45b780792f32.jpeg","isPro":false,"fullname":"Ohenenoo","user":"PeepDaSlan9","type":"user"},{"_id":"61f19829233c91cbd2f79e70","avatarUrl":"/avatars/a0735a94542b4f7cda5aed8bc4be0538.svg","isPro":false,"fullname":"Jian Ren","user":"alanspike","type":"user"},{"_id":"6248edbd483bb50494c8ac62","avatarUrl":"/avatars/58e8e4f303b644a2c35fd3a5e1fe54dc.svg","isPro":false,"fullname":"sbywqlq","user":"sbywqlq","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">

Papers

arxiv:2402.14797

Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis

Published on Feb 22, 2024

· Submitted by

AK on Feb 22, 2024

Upvote

Authors:

Aliaksandr Siarohin ,

Ivan Skorokhodov ,

Ekaterina Deyneka ,

Tsai-Shien Chen ,

Jian Ren ,

Abstract

Snap Video, a transformer-based model, addresses video generation challenges by extending EDM for spatial and temporal redundancy, achieving superior quality, consistency, and speed compared to existing U-Net based methods.

AI-generated summary

Contemporary models for generating images show remarkable quality and versatility. Swayed by these advantages, the research community repurposes them to generate videos. Since video content is highly redundant, we argue that naively bringing advances of image models to the video generation domain reduces motion fidelity, visual quality and impairs scalability. In this work, we build Snap Video, a video-first model that systematically addresses these challenges. To do that, we first extend the EDM framework to take into account spatially and temporally redundant pixels and naturally support video generation. Second, we show that a U-Net - a workhorse behind image generation - scales poorly when generating videos, requiring significant computational overhead. Hence, we propose a new transformer-based architecture that trains 3.31 times faster than U-Nets (and is ~4.5 faster at inference). This allows us to efficiently train a text-to-video model with billions of parameters for the first time, reach state-of-the-art results on a number of benchmarks, and generate videos with substantially higher quality, temporal consistency, and motion complexity. The user studies showed that our model was favored by a large margin over the most recent methods. See our website at https://snap-research.github.io/snapvideo/.