Podcast Atlas
concept Updated 2026-08-24 Topics: Technology, Culture

Live-Action Film Under AI

EP277 对话贾樟柯（下）：我没有背叛真实世界，我只是在寻找电影的新可能 adds 贾樟柯’s self-boundary. He is willing to test AI cinema and argues for Film Reality Beyond Realism, but says his own central interest remains the live-action practice of taking a camera into life scenes and capturing felt reality.

Live-action film under AI is the production question raised by 陆川 / Lu Chuan in E234｜未来实拍电影还存在吗？与导演陆川聊聊AI给影视人的恐惧与自由: when Video Models and AI Video Production Workflow can cheaply generate images, shots, and previs, which scenes still deserve human crews, locations, weather, cost, and physical shooting?

The source’s answer is not that live action disappears. AI becomes a measuring tool. Repetitive spectacle, generic coverage, and easily reproduced material may move to AI, while live action becomes more justified where real bodies, natural light, friendship, actor interaction, and unrepeatable emotional discovery matter.

《南京！南京！》 / Nanjing! Nanjing! supplies the source’s clearest example. Lu Chuan’s memory of finding a character’s execution-scene line with the actor on the shooting day defines live action as a process that can produce moments no one fully planned in advance.

从央视纪录片到爆款 AI 短剧：第一批「转身」的导演 | S10E11 adds the short-drama operator version. 抽象仔 / Chouxiangzai argues that AI is already appropriate for short-drama emotion because the format often needs large, direct, high-payoff feeling, while film-grade restraint such as a tear slowly forming and falling remains difficult and may justify hiring actors. 台台 / Taitai adds that long-form AI film also lacks a settled commercial path, not only a technical path.

Key Claims

  • Jia’s source adds that defending live action need not require rejecting AI; a director can test generated cinema while reserving his main artistic interest for photographed reality.
  • AI pressures creators to justify live-action cost scene by scene.
  • The strongest live-action defense is not technical fidelity but irreproducible human performance and presence.
  • AI may make physical production more luxurious and selective rather than obsolete.
  • Film education and in-person theater/film training still matter because embodied practice cannot be fully absorbed from model knowledge bases.
  • The concept complements Industrial-Grade Film Models: even better models increase the need to decide what should remain human-shot.
  • Short-drama suitability does not automatically imply feature-film readiness; production cycles, model iteration speed, budget, approval, and distribution all affect whether an AI film can close commercially.
  • AI performance may be good enough for broad short-drama emotions while still weaker for restrained, actor-driven emotional detail.

Connections