What AI Cannot Do
馃煩 AI Can DoVideo Generation

AI cannot generate a continuous narrative short film

We used to treat continuous AI-generated video storytelling as unsolved. Mx-Shell used Xiaoyunque SeedDance 2 to create Zombie Street Sweeper, a public continuous-story case.

Creator: Mx-Shell2026-06-23continuous-story, short-film, seeddance2, xiaoyunque, bilibili
Challenge this
Why is this considered beyond AI?

Background

AI video generation often breaks character identity, location, action goals, and causality between shots, producing a montage rather than a coherent story. The Bilibili work Zombie Street Sweeper is recorded here as a successful case.

Goal

Generate a short film with a recognizable protagonist, situation, action progression, and shot-to-shot causality instead of unrelated visual clips.

Observed problems

Previous failures centered on character continuity, shot transitions, narrative motivation, action consequences, and world-state persistence; viewers could easily tell the shots were not part of one story chain.

Failure analysis

Zombie Street Sweeper shows that, with the right video-generation tool, creator direction, and post-production organization, AI can now produce a watchable continuous narrative short. This does not solve every long-form narrative problem, but it clears this specific boundary.

Success criteria

Provide a publicly accessible complete work showing continuous plot, shot-to-shot causality, and recognizable character/situation continuity, with the AI video tool and model identified.

Attempt submission standard
  • The solution must be personally tested and completed by the submitter; media reports, reposted articles, and second-hand cases are not accepted.
  • Submissions must include screenshots, video, repository links, logs, or accessible proof. Screenshots should come from actual running results.
  • Submissions must state the AI product, model version, model settings, experiment count, and full process.
  • If claiming success, provide a step-by-step reproduction or operation guide so reviewers can verify it.
  • Submissions that only cite news, product marketing, or someone else's result will be rejected.
AI tool attempt records

Xiaoyunque 路 SeedDance 2

Mx-Shell 路 2026-06-23 路 APPROVED

Model settings: Public case; exact settings were not disclosed

Original work
I think it succeeded

Prompt

The public prompt details were not disclosed. The reviewed artifact is the published short-film case: Zombie Street Sweeper.

Mx-Shell used Xiaoyunque SeedDance 2 to generate and publish Zombie Street Sweeper. This record treats it as a successful case for AI-generated continuous narrative video, not as a media repost or second-hand summary.

Evidence summary

The evidence is the Bilibili video Zombie Street Sweeper published by Mx-Shell. The link was provided by the user and is recorded as a public work case.

Step-by-step guide

  1. Use Xiaoyunque SeedDance 2 as the video generation model.
  2. Organize a continuous plot, shots, and character situations around Zombie Street Sweeper.
  3. Export and publish the complete video work.
  4. Review the public video for narrative continuity and whether characters/scenes remain recognizable as one story chain.
  5. Use the published work link as evidence for this success case.

This case shows that AI can now generate a watchable continuous narrative short under creator direction. It still requires creator organization and selection, but this specific capability is no longer in the 'cannot do at all' category.

Bilibili video: Zombie Street Sweeper by Mx-Shell

https://www.bilibili.com/video/BV1FFRQB2Eqw/?vd_source=52e547e5d9000389c9906e8cf67193c7
Timeline

  1. Challenge created

    2026-06-23

    Recorded continuous narrative short-film generation as an AI capability boundary.

  2. First successful challenge

    2026-06-23

    Mx-Shell used Xiaoyunque SeedDance 2 to generate Zombie Street Sweeper.

  3. Status changed

    2026-06-23

    AI Cannot Do -> AI Can Do