{"paper":{"title":"Make-A-Video: Text-to-Video Generation without Text-Video Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A method turns text into videos by extending image generators with motion learned separately from unlabeled footage.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Adam Polyak, Devi Parikh, Harry Yang, Jie An, Oran Gafni, Oron Ashual, Qiyuan Hu, Sonal Gupta, Songyang Zhang, Thomas Hayes, Uriel Singer, Xi Yin, Yaniv Taigman","submitted_at":"2022-09-29T13:59:46Z","abstract_excerpt":"We propose Make-A-Video -- an approach for directly translating the tremendous recent progress in Text-to-Image (T2I) generation to Text-to-Video (T2V). Our intuition is simple: learn what the world looks like and how it is described from paired text-image data, and learn how the world moves from unsupervised video footage. Make-A-Video has three advantages: (1) it accelerates training of the T2V model (it does not need to learn visual and multimodal representations from scratch), (2) it does not require paired text-video data, and (3) the generated videos inherit the vastness (diversity in ae"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the proposed spatial-temporal decomposition and pipeline can transfer motion dynamics learned from unsupervised video to text-conditioned generation without introducing visible artifacts or losing text faithfulness.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Make-A-Video achieves state-of-the-art text-to-video generation by decomposing temporal U-Net and attention structures to add space-time modeling to text-to-image models, trained without any paired text-video data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A method turns text into videos by extending image generators with motion learned separately from unlabeled footage.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6246920416a3dc97bd69ce776b67afac4717b24d3e8991f1286e9eea16686c2e"},"source":{"id":"2209.14792","kind":"arxiv","version":1},"verdict":{"id":"b692e49f-8022-4dc6-8d28-ef916e606e9c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T01:08:48.153334Z","strongest_claim":"Make-A-Video sets the new state-of-the-art in text-to-video generation, as determined by both qualitative and quantitative measures.","one_line_summary":"Make-A-Video achieves state-of-the-art text-to-video generation by decomposing temporal U-Net and attention structures to add space-time modeling to text-to-image models, trained without any paired text-video data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the proposed spatial-temporal decomposition and pipeline can transfer motion dynamics learned from unsupervised video to text-conditioned generation without introducing visible artifacts or losing text faithfulness.","pith_extraction_headline":"A method turns text into videos by extending image generators with motion learned separately from unlabeled footage."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2209.14792/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":15,"sample":[{"doi":"","year":2005,"title":"Language Models are Few-Shot Learners","work_id":"214732c0-2edd-44a0-af9e-28184a2b8279","ref_index":2,"cited_arxiv_id":"2005.14165","is_internal_anchor":true},{"doi":"","year":null,"title":"arXiv preprint arXiv:2204.14217 , eprint =","work_id":"1768d36e-3377-4973-b3f4-698595fdb671","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Make-a-Scene","work_id":"2c99d6c4-d1c6-4714-95e0-006639b9ac5c","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1109/cvprw50498.2020.00193","year":2020,"title":"Score-cam: Score-weighted visual explanations for convolutional neural net- works","work_id":"cafb0da8-3304-4402-975f-3fe00918e3fa","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2006,"title":"Denoising Diffusion Probabilistic Models","work_id":"dc023f4e-7c79-471c-b713-deeb559ba010","ref_index":6,"cited_arxiv_id":"2006.11239","is_internal_anchor":true}],"resolved_work":15,"snapshot_sha256":"afc3ac3c5164d2f43c2309612ee8c8c49567e878695321679744c868f5b96609","internal_anchors":12},"formal_canon":{"evidence_count":2,"snapshot_sha256":"4d1ceaf666a386c984c02a6817e4891b399ef4a29719fb472265af1d22699866"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}