0:00 / 0:00

Erik Gen

@erik_gen_ai

INSTAGRAM · 2026-01-28Source

69likes

7comments

Remix This

Recreate with Seedance 2

Make your own AI viral video

Prompt

GLOBAL LOCK: A middle-aged man with salt-and-pepper hair and a short beard, wearing a black zip-up hoodie and grey jeans. He drives a red modern SUV. The visual style is cinematic photorealism, shot on 35mm film with subtle grain and high dynamic range. The environment is Chicago, featuring recognizable landmarks and street signs.

[00:00–00:17] Nighttime city driving. Close-up of the man's face illuminated by warm dashboard lights and passing orange streetlamps. Medium shot from the backseat looking over his shoulder at the steering wheel and the blurred city traffic ahead. Smooth gimbal movement.

[00:18–00:44] Driving through a long, dimly lit concrete tunnel. Rhythmic orange sodium lights flash across the car's interior. The man looks tired, rubbing his eyes and face. POV shot through the windshield of the red SUV following another car. The sound of the engine echoes.

[00:45–01:18] The car emerges from the tunnel into a desaturated, foggy daylight. The city is abandoned. Overgrown vines on buildings, cracked pavement, and no other people. The man looks around in shock and confusion. Wide establishing shots of the empty city.

[01:19–02:41] Wide shots of the red SUV driving slowly through empty, debris-strewn Chicago streets. Close-up of the man looking out the window. Graffiti on a wall reads "THE END IS NEAR". The camera tracks the car from a low angle.

[02:42–04:31] The man gets out of the car on a bridge overlooking a river. He walks along the railing. Suddenly, massive skyscrapers in the background begin to crumble and collapse into the water with huge splashes of dust and debris. He covers his face in terror. High-angle drone shots of the destruction.

[04:32–04:59] Sudden cut back to the night tunnel. The red SUV is smashed in a horrific accident, smoke rising from the hood. Paramedics are performing CPR on the man lying on the ground. They place an oxygen mask on him. He opens his eyes wide in a final gasp. The ambulance drives away into the dark tunnel with flashing red and blue lights.

NEGATIVE PROMPT: cartoon, anime, 3d render, low resolution, distorted faces, inconsistent car model, bright happy colors in the abandoned city, text logos on clothing, flickering, morphing limbs, unnatural movements.

SOUND DESIGN PACK:
[00:00] Muffled city traffic, low synth drone, ticking clock sound.
[01:18] Transition sound: a sharp metallic ring fading into eerie wind noise.
[04:13] Loud rumbling, crashing concrete, splashing water, deep bass drops.
[04:32] High-pitched ringing, distant sirens, rhythmic thumping of chest compressions, heavy breathing.

How erik_gen_ai Made This Cinematic AI Short Film Post Apocalyptic Twist Video — and How to Recreate It

This case study analyzes a high-production-value AI short film titled "a film by Erik Gen." The video is a masterclass in cinematic narrative consistency, blending a "liminal space" aesthetic with a psychological thriller twist. It features a middle-aged man driving a red SUV through a vibrant, nighttime Chicago, which seamlessly transitions into a haunting, post-apocalyptic, foggy version of the same city. The core keywords for this aesthetic are cinematic photorealism, abandoned urban exploration, post-apocalyptic Chicago, and psychological narrative twist. By maintaining strict character and vehicle consistency across nearly five minutes of footage, the creator demonstrates the current ceiling of AI video generation.

What You’re Seeing: A Visual Analysis

The video opens with high-contrast night shots. We see a man with salt-and-pepper hair and a trimmed beard, wearing a black zip-up hoodie. The lighting is motivated by the dashboard's glow and passing orange streetlights. The texture is distinctly filmic, with a 35mm grain and realistic motion blur.

As the man enters a tunnel, the rhythm of the editing matches the flashing of the overhead lights. The transition to the "abandoned city" is marked by a shift in color grading—from warm, saturated oranges and reds to a cold, desaturated, and foggy grey palette. The environment is rich with detail: cracked asphalt, vines climbing skyscrapers, and "The End is Near" graffiti. The climax features complex physics simulations of buildings collapsing into a river, followed by a jarring return to reality—a car crash scene illuminated by the red and blue strobes of emergency vehicles.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:17	Man driving SUV in night traffic.	CU / Over-the-shoulder	Warm, high-contrast night	Establish realism and character.
00:18–00:44	Driving through a concrete tunnel.	POV / Interior MS	Rhythmic orange flicker	Build tension and transition.
00:45–01:18	Emerging into abandoned, foggy city.	Wide / Establishing	Desaturated, cold grey	The "Hook": Mystery reveal.
01:19–03:00	Exploring empty Chicago streets.	Tracking / Low angle	Muted, overcast	World-building and immersion.
03:01–04:31	Buildings collapse into the river.	Wide / High angle	Gloomy, dusty	Spectacle / Climax.
04:32–04:59	Car crash reveal / Paramedics.	Top-down / Extreme CU	Red/Blue emergency strobes	Emotional payoff / Twist.

Why It Went Viral: The Narrative Hook

The Power of the "Liminal Space"

The video taps into the "liminal space" and "abandoned world" tropes that are currently dominating internet culture. Seeing a familiar, bustling city like Chicago completely empty triggers a biological sense of "uncanny valley" curiosity. It forces the viewer to ask, "Where is everyone?" which significantly boosts retention.

The "Near-Death Experience" Trope

Psychologically, the "it was all a dream/hallucination" twist is a classic narrative device. By framing the post-apocalyptic world as a dying man's vision during a car crash, the creator adds a layer of emotional weight that simple "AI art" usually lacks. This transition from sci-fi spectacle to human tragedy encourages re-watching to find "clues" in the earlier scenes.

Extreme Technical Consistency

For indie creators, the biggest hurdle is keeping the character and car the same across shots. This video succeeds where most fail. The red SUV and the man's facial features remain stable for 5 minutes, which signals "high quality" to the viewer and the platform's algorithm, separating it from low-effort AI "slop."

Platform Perspective & Signals

On platforms like Instagram and TikTok, watch time is king. The 5-minute duration is risky, but the constant progression of the environment (Night -> Tunnel -> Fog -> Collapse -> Crash) provides a "micro-hook" every 15-20 seconds. The "Save" rate is likely high because other creators want to study the consistency prompts, and the "Share" rate is driven by the "Did you see the ending?" factor.

5 Testable Viral Hypotheses

The Contrast Hook: If you start with extreme realism and transition to surrealism, you capture both the "tech" and "art" audiences.
The Landmark Effect: Using a recognizable city (Chicago) increases shareability among locals and people familiar with the geography.
The Rhythmic Tunnel: Using rhythmic light patterns (like the tunnel sequence) creates a hypnotic effect that increases early-video retention.
The Silent Narrative: Removing dialogue and relying on cinematic sound design reduces the "language barrier," making the video globally viral.
The Twist Payoff: A final 5-second reveal that recontextualizes the previous 4 minutes will always drive higher comment section engagement.

How to Recreate: From 0 to 1

Step 1: Character & Asset Locking

Use a tool like Midjourney to create a "Character Sheet" of your protagonist. For this video, the prompt would focus on "Middle-aged man, salt-and-pepper hair, black zip-up hoodie." Save this image and use it as a Character Reference (--cref) in your video generation tool.

Step 2: Vehicle Consistency

Similarly, generate a specific car (e.g., "Red modern SUV, sleek design") and use it as an Image Reference for every driving shot. Consistency is the difference between a "video" and a "film."

Step 3: The Night Sequence

Generate shots of the man driving at night. Focus on lighting prompts: "cinematic lighting, dashboard glow, orange streetlights reflecting on glass."

Step 4: The Transition (The Tunnel)

Create a sequence in a concrete tunnel. Use motion prompts like "fast camera movement, rhythmic flashing lights" to bridge the two worlds.

Step 5: The Post-Apocalyptic World

Generate the "Day" version of your city. Use keywords: "overgrown, abandoned, heavy fog, cracked roads, desaturated color grade." Ensure the red SUV is still present.

Step 6: Physics & Spectacle

For the collapsing buildings, use video-to-video tools or specific physics-based prompts in Runway Gen-3: "skyscraper crumbling into river, massive dust clouds, water splashes, slow motion."

Step 7: The Twist Reveal

Film or generate the "Real World" scene. Use high-intensity lighting (emergency strobes) to create a jarring contrast with the previous foggy scenes.

Step 8: Sound Design & Editing

Do not use generic music. Layer sounds: low engine hums, wind whistling through empty streets, the sound of crumbling stone, and finally, the sharp sound of sirens.

Growth Playbook: Distribution Strategy

3 Opening Hook Lines

"I drove through a tunnel and the world changed..."
"The ending of this AI film will break your heart."
"How I maintained 100% character consistency for 5 minutes."

Caption Templates

The Storyteller:
What if the world you see isn't the one you're in? 🌫️ I spent weeks perfecting the consistency for this short film. Watch until the end to see the truth. #AIShortFilm #CinematicAI #Storytelling

The Tech Breakdown:
Consistency is the hardest part of AI video. 🚗 Here is how I kept this red SUV and character identical across 50+ shots. Which scene was your favorite? #RunwayGen3 #Midjourney #AITutorial

Hashtag Strategy

Broad: #AI #ShortFilm #Cinematic #DigitalArt
Mid-tier: #PostApocalyptic #LiminalSpaces #ChicagoCreators #VisualEffects
Niche: #RunwayGen3 #AIStorytelling #ErikGen #AIFilmFestival

Frequently Asked Questions

What tools make it look the most similar?

Midjourney for character/car references and Runway Gen-3 or Luma Dream Machine for the video generation.

What are the 3 most important words in the prompt?

"Cinematic," "Photorealistic," and "Consistency."

Why does the generated face look inconsistent?

You likely aren't using a Character Reference (cref) or a fixed seed across your generations.

How can I avoid making it look like AI?

Add film grain in post-production and focus on "motivated lighting" rather than flat, bright scenes.

Is it easier to go viral on Instagram or TikTok with this?

Instagram's algorithm currently favors high-fidelity, cinematic "aesthetic" content more than TikTok's fast-paced UGC style.