By early 2028 (January 1, 2028), will an AI system be able to generate a full high-quality movie from a text prompt — e.g. 'make me a 120-minute Star Trek / Star Wars crossover' — that is more or less comparable to a big-budget studio film in production quality (visuals, coherent narrative, character consistency, sound/score), even if it doesn't fully pass a Turing Test? The AI does not need to be publicly available — it just needs to be confirmed to exist (e.g. via credible demonstration, research publication, or announcement by a major AI lab or studio).

probability: 11%

rationale:

To achieve a YES resolution by January 1, 2028, an AI system must satisfy a highly demanding conjunction of requirements: autonomous generation from a text prompt, a runtime of at least 90 minutes, and quality broadly comparable to a big-budget studio release (including coherent plot, consistent characters, pacing, and professional audio). Based on the trajectory of AI video capabilities as of mid-2026, meeting all of these criteria simultaneously within the next 18 months remains very unlikely.

Current native single-generation clips from frontier systems are still largely measured in seconds. For instance, Microsoft's March 2026 Sora 2 documentation specifies 1–20 seconds at up to 1080p learn.microsoft.com, Google DeepMind's Veo 3.1 benchmark videos sit around 8 seconds deepmind.google, and competitors like Kling 3.0 operate in the ~15-second range lushbinary.com. While systems like Runway Gen-4 are making progress on character and object consistency from reference images runwayml.com, bridging the gap from minute-scale generation to a coherent 90-minute narrative represents a massive leap in temporal coherence and multi-scene planning.

Feature-length AI films have begun to emerge, but they clearly illustrate the distance remaining to the studio-quality threshold. Projects like 'Hell Grind' (95 minutes) and 'Dreams of Violets' (75 minutes) rely heavily on human curation, requiring thousands of manual generations and heavy editing cined.comvariety.com. Furthermore, they fall far short of the required quality bar; reviews note 'nonsensical' plots , jitter, weak choreography, awkward dialogue, and inconsistent accents joblo.com. Even well-funded efforts like the OpenAI-backed 'Critterz' are highly human-led productions rather than autonomous single-prompt generations variety.comdeadline.com.

There is some upside potential that warrants a non-trivial probability. The next 18 months could plausibly yield an agentic production pipeline—where an LLM handles screenwriting and planning, orchestrating short-clip video models, voice, and music generation to internally assemble a feature from a single prompt. A major lab could conceivably produce a credible, closed demonstration of such a system, satisfying the resolution criteria without full public release. Major investments, such as Google's ~$75M partnership with A24 variety.com, indicate that capital and industry focus are aggressively targeting long-form AI video.

However, expecting an autonomous system to match the narrative coherence, emotional resonance, zero-defect continuity, and precise editing rhythm of a major studio film without human curation is an exceptionally steep hurdle. Additionally, commercial incentives heavily favor human-in-the-loop tools for filmmakers rather than full text-to-movie generation. Given the qualitative leaps required in acting, dialogue, and physics, a fully prompt-generated, 90-minute studio-comparable film is highly improbable by early 2028.