Ibrahim Diallo describes a very relatable “prompt-to-film” dream: take an old short story, feed a script and sketches into Sora (or Runway/Veo), and get something that looks like the movie in your head. The reality, he argues, is that even when the clips look impressive in isolation, they’re hard to steer toward a specific narrative. The models produce scenes that are “adjacent” to what you asked for: visually plausible, but generic, cliché, and inconsistent in the ways that matter for continuity.
He frames this as an emerging genre with its own telltale look: “AI videos” have an aesthetic fingerprint (a subtle wrongness) that people can recognize even when they can’t articulate why. That’s not just a taste issue — it changes where these tools work well. If audiences develop an instinctive aversion to the style, then “serious” creative uses are fighting an uphill battle.
Where the tools do thrive is in high-volume manipulation: spam, scams, rage-bait, and misinformation (often aimed at older adults), where consistency and intent are less important than producing convincing-looking clips cheaply and endlessly. It’s a sharp reminder that the biggest near-term impact of generative video might be less about filmmaking and more about content abuse.