26th June
Videos generated via Davinci.at, using the model seeddance 2.0
Prompt:
‘I want a video of a woman sat in a modern kitchen at a table. I want her to pick up a glass of water and drink it. I want it repeated in the following styles:
Action, Documentary, Sci-fi, Horror, Noir, Romance’
Video 1.
 | Frame 1 (golden warm light, shallow focus, intimate close-up) → Sofia Coppola — that dreamy, warm, slightly melancholic closeness to a woman’s face is her signature. | Still from Sofia Coppola’s Priscilla, 2023 |
 | Frame 6 (low table angle, flat daylight, neutral) → Yorgos Lanthimos — that uncomfortably low, wide-angle perspective with detached coldness is very much his thing (The Favourite, Dogtooth). | Still from Yorgos Lanthimo’s Dogtooth, 2009 |
 | Frame 5 (wider shot, clean naturalistic light, calm) → Kenneth Lonergan — understated, observational, feels like a quiet domestic drama moment. | Still from Kenneth Lonergan’s You can Count on Me, 2000 |
 | Frame 4 (cool blue underlighting, glowing glass, clinical) → David Fincher — the cold precision, the underlighting, the sense of dread in a mundane setting. Very Gone Girl. | Still from David Fincher’s Gone Girl, 2014 |
 | Frame 3 (extreme B&W contrast, wide fearful eyes, horror energy) → Robert Eggers — that expressionistic, dread-soaked close-up screams The Witch or Nosferatu. | Still from Robert Eggers’ Nosteratu, 2024 |
 | Frame 2 (black and white, harsh chiaroscuro, noir) → Pablo Larraín — high contrast B&W with an almost suffocating intimacy, reminiscent of Spencer or Jackie. | Still from Pablo Larrain’s Jackie, 2016 |
Video 2.
 | Frame 7 (flat, neutral, diffuse light, centred framing) → Genre: Realist drama / stock footage baseline → Director: Chloé Zhao — that unaffected, documentary-adjacent stillness, natural light, person just existing in a space. | Still from Chloe Zhao’s Nomadland 2020 |
 | Frame 6 (smoke/fog billowing, blinding backlight, almost surreal) → Genre: Surrealist art cinema / horror → Director: Gaspar Noé — that overwhelming, hallucinatory atmosphere with the blinding central light is very Enter the Void or Climax. | Still from Gaspar Noe’s Climax, 2019 |
 | Frame 5 (holographic UI overlays, desaturated, sci-fi world-building) → Genre: Cyberpunk / speculative fiction → Director: Ridley Scott — those layered holographic interfaces feel very Blade Runner original. | Still from Ridley Scott’s Blade Runner, 1982 |
 | Frame 4 (hyper-clean, bleached whites, glowing futuristic counter) → Genre: Sci-fi / tech thriller → Director: Alex Garland — sterile, clinical, beautiful unease. Pure Ex Machina. | Still from Alex Garland’s Ex Machina, 2014 |
 | Frame 3 (B&W, shadowy figure looming behind her, menacing) → Genre: Psychological horror / thriller → Director: David Fincher — the unseen threat in the background, the controlled darkness, that sense of dread. Very Zodiac or Se7en | Still from David Fincher’s Zodiac, 2007 |
 | Frame 2 (B&W, same blind shadows, more contrast) → Genre: Classic noir → Director: Steven Soderbergh — that sharp, graphic B&W with hard shadow geometry has his cool, controlled hand all over it (The Knick, Logan Lucky). | Still from Steven Soderbergh’s Logan Lucky |
 | Frame 1 (warm golden sun, dramatic venetian blind shadows striping the background) → Genre: Neo-noir / psychological thriller → Director: Denis Villeneuve — those bold geometric shadow patterns feel very Blade Runner 2049 or Sicario. Gorgeous but ominous. | Still from Denis Villeneuve’s Sicario, 2015  573 × 237 |
Good observation! Honestly, a few reasons:
The AI model itself skews contemporary — the aesthetic defaults of current video AI (clean kitchens, white shirts, shallow focus, moody underlighting) are trained heavily on modern commercial and streaming content, so the visual language it produces naturally echoes directors who are prominent now.
The styles it can replicate are ones with lots of training data — Fincher, Villeneuve, Garland etc. have highly distinctive, widely discussed and documented visual signatures that are well represented online, which is exactly what AI models learn from.
Noir and classic cinema is harder to prompt — older directors like Kubrick, Cassavetes, Tarkovsky, Bergman, or Ophüls had styles rooted in film grain, anamorphic lenses, and practical lighting that AI video currently struggles to authentically reproduce. When you get B&W it tends to look like contemporary B&W (clean, digital) rather than genuine 1950s or 1960s celluloid.
I was also pattern-matching to what felt plausible — since this is AI-generated footage, it made sense to name directors whose work the model might plausibly have “absorbed.”
If you wanted to push toward older directors — say Kubrick’s symmetry, or Fellini’s chaos, or Bresson’s austerity — it would be interesting to see how hard you’d have to work against the model’s defaults to get there. Have you tried prompting for specific older styles?
Interesting comparison’s – some are a bit wide of the mark in terms of accurately connecting them with a director, but mostly genre remains accurate. In terms of the references, it’s interesting that (in particular the BW films) Hitchcock, Carol Reed, even David Lynch for the more surreal ones.
I don’t know yet what this has told me. Mainly that the models being used for image generation are probably/possibly skewing towards newer, more prevalent styles and media. This may mean that the influence of individuals, or of older media may not be acknowledged, and the canon of influence may not be seen or mapped. In comparison to older versions of appropriation, where the work is ripped and kind of acknowledged. Does this mean it’s not being critical, influenced by something generated obfuscates the lineage and critical position of a movement?
Discussed this with friends, we spoke about knowing the provenance of a film. It turned to epistological arguaments – what’s the difference between machine knowing and individuals knowing.