Passenger

Inspiration

Passenger grew out of a simple nightmare image: waking up on a train that never stops, with no memory and no way off. We wanted to mix puzzle logic, liminal horror and surreal rules—each carriage as a new “level” where physics, time or social norms are broken.

What it does

Passenger is a psychological horror short set on an endless train running through a ruined city. Each carriage is its own trial: gravity flips, time loops, passengers speak in riddles, and a faceless pursuer adapts to the hero’s fears and mistakes.

How we built it

We started from the logline and a list of “rules” for each carriage: one breaks gravity, one breaks time, one turns passengers into a chorus, etc. That gave us a clear structure for shots and transitions.

For visuals, we used AI image tools to prototype the train interior, ruined city exteriors and the look of the faceless pursuer. We then generated and refined moving shots with a mix of AI video tools (such as KLING, VEO, WAN, LTX Studio, SORA2, Luma, Pixverse and Runway).

Editing and pacing were done in CapCut, with Topaz Upscale for cleanup when needed. Voices, whispers and sound design were created and layered using AI TTS and SFX tools like ElevenLabs, ThinkSound and MiniMax TTS to keep the soundscape tense but not overwhelming.

Challenges we ran into

Consistency in a shifting world – keeping the train recognizably the same space while every carriage changes its rules. Designing the faceless pursuer – making it unsettling and readable in 9:16 without relying on gore. Pacing horror vertically – timing reveals, loops and jumps so they work both as standalone clips and as a continuous journey. Managing many tools – coordinating different models, resolutions and quirks into a single coherent pipeline.

Accomplishments that we're proud of

We created a clear “rule-based” horror world where each carriage feels distinct but part of one nightmare system.

We found a visual language for the faceless pursuer that stays creepy in close-ups, silhouettes and wide shots.

We built a reusable AI workflow for surreal, environment-driven horror that can extend beyond this one train story.

What we learned

We learned how important it is to define rules first in puzzle-horror: if the audience understands the pattern, every break in that pattern hits harder. We also saw how powerful AI tools are for iterating on impossible spaces—but only when we give them specific, shot-like directions instead of vague prompts. Designing for 9:16 forced us to rethink how to stage fear: vertical framing changes how bodies, doors and corridors read on screen.

What's next for Passenger

Next steps include expanding Passenger into a sequence of short “car episodes,” each built around a new set of rules and a new way the train reacts to the protagonist. We want to explore interactive possibilities—polls or branching cuts where viewers choose which door or carriage he takes next.