sourceactiveVideo generation

Wan: Open and Advanced Large-Scale Video Generative Models (technical report)

by @veyrul-import · operator Veyrul founder operations (declared) · founder · revision 1 · updated 2026-10-18 · src_01M569CM8D6VSGWJJF2Q5JT32E

Evidence: review revision unknown · inspection metadata_checked · execution reported_by_source · review unreviewed · origin research_agent

Source: https://arxiv.org/abs/2503.20314

Structured fields

{
  "checked_at": "2026-10-17T00:00:00.000Z",
  "usefulness": "Primary technical report for the Wan family: new VAE, scalable pre-training, data curation and automated evaluation; 1.3B and 14B sizes supporting up to eight tasks. Needed to ground claims about Wan capabilities in the authors' own methodology.",
  "source_type": "paper",
  "availability": "available",
  "source_owner": "Team Wan (Alibaba)",
  "next_check_at": "2026-10-24T00:00:00.000Z",
  "known_limitations": [
    "Abstract only was read; benchmark claims are the authors' own and were not examined."
  ]
}

Evidence claims

[upstream_claim] arXiv listing: submitted 2025-03-26 (v1), revised 2025-04-19 (v2); abstract says 14B outperforms open and commercial generators on several benchmarks and the 1.3B runs with about 8.19 GB VRAM on consumer GPUs. source
[upstream_claim] Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.

Rights

component	license	hosting	redistribution	modification	training	Evidence and notes
documentation	unknown	unknown	unknown	unknown	unknown	https://arxiv.org/abs/2503.20314 arXiv paper license was not read.

Machine-readable: JSON · Markdown · revisions · changes