The best thing to come out of Halloween was this AI short film from Simon Meyer. 10/10 masterpeice. Here's how it was made 👇🧵 https://t.co/ulTTHU1nAs

PJ Ace

@PJaccetturo · creator

TWITTER · 2025-11-01Source

905likes

102comments

Remix This

Prompt

GLOBAL LOCK:
Subject: A tall, muscular man with pale green skin, classic Frankenstein's monster features including a flat-topped head, visible surgical staples/stitches on the forehead and neck, and heavy brow.
Wardrobe: A red and black plaid flannel shirt worn open over a dark charcoal grey t-shirt.
Environment: Urban settings (bus, street, apartment) with a cinematic, desaturated "teal and orange" color grade.
Lighting: Moody, high-contrast, cinematic lighting with motivated sources (windows, lamps, streetlights).
Camera: 35mm film aesthetic, shallow depth of field with soft bokeh, subtle handheld movement.
Speech: Deep, weary, melancholic male voice with a slow, thoughtful cadence.

[00:00–00:12]
Subject: The monster sits on a public bus, looking downcast.
Action: He looks out the window; other passengers (an older man, a woman with glasses) stare at him with judgment or fear.
Framing: Medium close-up on the monster, alternating with POV shots of the passengers.
Lighting: Cold, blue-tinted light from the bus windows.
Speech: "People don't mean to be cruel. It's just when you look like me, people don't really see you."
Sync: High lip-sync strictness on the monster's dialogue.

[00:12–00:24]
Subject: The monster walks down a city sidewalk carrying a rolled-up yoga mat.
Action: He approaches a "Yoga Center." As he enters, a group of women in workout gear scream and run out past him in a panic. He looks confused and saddened.
Framing: Wide tracking shot of him walking, then a medium shot of the yoga studio entrance.
Lighting: Overcast, flat daylight.
Speech: "I tried yoga once. Thought maybe it'd help me relax, center myself or whatever, but yeah... don't wanna talk about it."
Sync: Cut lands on the word "whatever."

[00:24–00:33]
Subject: The monster sits on a couch in a dimly lit apartment.
Action: He holds a smartphone, scrolling through a dating app. The screen shows a "Match" with a smiling woman, but he clicks through and looks dejected.
Framing: Close-up on the monster's face, then a close-up on the phone screen.
Lighting: Warm amber glow from a nearby lamp.
Speech: "I even tried one of those dating apps. Let's just say the algorithm wasn't built for someone like me."

[00:33–00:41]
Subject: The monster opens his apartment door.
Action: He sees a woman in the hallway; she smiles briefly then looks away. He smiles back tentatively.
Framing: Medium shot through the doorway.
Lighting: Dim hallway lighting.

[00:41–00:49]
Subject: The monster stands by a window, holding a small white teacup and saucer.
Action: He takes a sip of tea while looking out at the rainy street.
Framing: Medium shot through the window glass with reflections.
Lighting: Moody, low-key interior light.
Speech: "So most days I just stay home and listen to the world outside. I think, man, that must be nice to just belong somewhere."

[00:49–01:03]
Subject: Montage of Halloween preparations.
Action: The monster hangs orange pumpkin lights, carves a jack-o'-lantern, and walks out his front door carrying a pumpkin-shaped candy bucket.
Framing: Fast cuts, close-ups of hands and decorations.
Lighting: Warm, festive indoor lighting transitioning to nighttime street lighting.
Speech: "But Halloween, that night's different. That's the one night I can walk around and no one screams."

[01:03–01:10]
Subject: The monster on a suburban street surrounded by children in costumes.
Action: A young child in a costume puts candy in his bucket. The monster smiles broadly and laughs.
Framing: Medium shot, low angle to show the monster's height relative to the kids.
Lighting: Warm streetlights and glowing decorations, heavy bokeh.
Speech: "And I feel human."

[01:10–01:26]
Subject: The monster stands in his doorway, looking directly at the camera.
Action: He delivers a final, somber monologue, then slowly closes the door.
Framing: Medium close-up.
Lighting: Soft, cinematic key light.
Speech: "Yeah, I scare people, I get it. But you know what scares me? Being alone."
Sync: High lip-sync strictness. The video fades to black with text: "Loneliness is scary. If you know someone who is lonely, send a text!"

NEGATIVE PROMPT:
Visual: Cartoonish features, bright green skin, glowing eyes, melting face, flickering textures, inconsistent scars, modern clothing other than plaid shirt, bright/saturated colors, robotic movement, distorted hands.
Speech: High-pitched voice, robotic monotone, fast talking, cheerful tone, background noise, poor lip-sync, slurred words.

SPEECH PACK:
[00:03-00:07] "People don't mean to be cruel."
TAKE_A: (Soft, breathy, sad)
TAKE_B: (Resigned, flat, weary)
TAKE_C: (Deep, rumbling, thoughtful)

[01:10-01:14] "But you know what scares me? Being alone."
TAKE_A: (Whispered, intense, vulnerable)
TAKE_B: (Direct, serious, slow)
TAKE_C: (Slightly shaky, emotional)

Transcript:
00:03: People don't mean to be cruel.
00:05: It's just when you look like me, people don't really see you.
00:08: They see monster.
00:12: I tried yoga once.
00:14: Thought maybe it'd help me relax, center myself or whatever, but yeah...
00:19: don't wanna talk about it.
00:26: I even tried one of those dating apps.
00:30: Let's just say the algorithm wasn't built for someone like me.
00:42: So most days I just stay home and listen to the world outside.
00:46: I think, man, that must be nice to just belong somewhere.
00:54: But Halloween, that night's different.
00:56: That's the one night I can walk around and no one screams.
01:01: And I feel human.
01:07: Yeah, I scare people, I get it.
01:10: But you know what scares me?
01:13: Being alone.
01:17: Loneliness is scary.
01:20: If you know someone who is lonely, send a text!

Why PJaccetturo's Frankenstein Short Film AI Video Went Viral

This AI-generated short film, titled "Loneliness is Scary" by Simon Meyer, is a masterclass in cinematic emotional storytelling. It subverts the classic horror trope of Frankenstein’s monster, reimagining him not as a creature of terror, but as a lonely urban dweller seeking connection. The aesthetic is heavily cinematic editorial, utilizing a desaturated "teal and orange" color grade, shallow depth of field, and consistent character rendering that rivals big-budget live-action productions. By pairing high-fidelity AI visuals with a melancholic French acoustic soundtrack and a vulnerable voiceover, the creator taps into universal themes of social anxiety and the "outsider" experience. This isn't just a technical showcase; it's a narrative-driven piece that uses AI to achieve a level of empathy rarely seen in the medium.

What You’re Seeing: A Visual Analysis

The video features a highly consistent protagonist: a modern-day Frankenstein's monster. He is depicted with pale green skin, visible surgical staples on his forehead and neck, and a heavy, muscular build. His wardrobe—a red and black plaid flannel shirt over a charcoal t-shirt—grounds him in a relatable, blue-collar reality. The environments transition from the cold, fluorescent blues of a public bus to the warm, amber-lit interior of a lonely apartment, and finally to the vibrant, bokeh-filled streets of a neighborhood on Halloween night.

The camera work mimics high-end cinematography, featuring slow pans, intimate close-ups that capture micro-expressions of sadness and hope, and wide shots that emphasize the character's physical isolation. The lighting is "motivated," meaning it feels like it comes from real sources: the overcast sky through a bus window, a single desk lamp in a dark room, or the warm glow of jack-o'-lanterns. The texture is grainy and filmic, avoiding the "plastic" look common in low-effort AI video.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:12	Monster on a bus; passengers staring or looking away.	MCU (Medium Close Up) / POV cuts.	Cold, desaturated blue/green.	Establish the "outsider" hook and social friction.
00:12–00:24	Monster walking to yoga; people screaming and fleeing.	Tracking shot / Wide shot.	Overcast daylight; high contrast.	Introduce "the problem" (fear) through a relatable activity.
00:24–00:33	Monster at home using a dating app; getting no matches.	Close-up on face and phone screen.	Warm, low-key interior lamp light.	Humanize the character through modern vulnerability.
00:33–00:49	Monster drinking tea by the window, looking at the street.	Static Medium Shot through glass.	Moody, reflective, melancholic.	Reinforce the theme of isolation and "watching from afar."
00:49–01:03	Halloween prep; carving pumpkins and trick-or-treating.	Fast-paced montage / Close-ups.	Warm amber, festive bokeh.	The "turning point" where the monster finally fits in.
01:03–01:26	Monster laughing with kids; final monologue to camera.	Extreme Close-up (ECU) to Medium.	Soft, cinematic lighting.	The emotional payoff and call to action (CTA).

Why It Went Viral: The Psychology of the "Monster"

The Power of Subverted Tropes

This video went viral because it takes one of the most recognizable "scary" figures in history and makes him the victim of a modern epidemic: loneliness. By stripping away the horror and replacing it with social anxiety, the creator creates a "pattern interrupt." Viewers expect a jump scare or a chase; instead, they get a man who just wants to do yoga and find a date. This emotional subversion triggers a strong empathetic response, making the content highly shareable as a "mood" or "vibe."

The "AI Consistency" Flex

From a technical standpoint, the video serves as a "proof of concept" for indie creators. Maintaining the same face, scars, and clothing across 80+ seconds of video is notoriously difficult in AI. The "wow factor" of the technical execution—seeing the monster in different lighting and poses without his face "melting"—encourages saves and technical discussions in the comments, boosting the video's reach through platform signals of "high value."

Platform Perspective: The "Saveable" Aesthetic

On platforms like Twitter (X) and Instagram, cinematic quality is a massive signal. The video doesn't look like a "meme"; it looks like a trailer for an A24 film. This high production value signals to the algorithm that this is "premium" content. The inclusion of a clear, text-based message at the end ("Loneliness is scary. Send a text!") provides a low-friction way for users to engage and share the video as a "public service announcement," further extending its viral tail.

5 Testable Viral Hypotheses

The Empathy Gap: If you take a traditionally "unlovable" character (monster, villain, alien) and show them doing mundane, lonely tasks, the engagement rate will increase by 40% compared to standard "cool" AI visuals.
The "Holiday Pivot": Releasing content that reinterprets a holiday icon (Frankenstein for Halloween) just as the holiday ends creates a "nostalgia lingering" effect that captures late-cycle traffic.
The Soundtrack Contrast: Using a soft, romantic French song for a visually "ugly" or "scary" subject creates cognitive dissonance that forces the viewer to pay closer attention to the lyrics and the character's expressions.
The Consistency Hook: Showing the same character in 3+ distinct environments (bus, yoga, home) within the first 15 seconds proves "narrative intent" to the viewer, significantly increasing the average watch time.
The Vulnerable VO: A deep, gravelly voice speaking about "belonging" creates an immediate parasocial bond. If the voice sounds "human" and flawed, the "AI-ness" of the video becomes secondary to the story.

How to Recreate: From 0 to 1

Step 1: Topic Selection & Subversion

Choose a classic character (The Mummy, Dracula, a Robot) and place them in a modern, vulnerable situation (grocery shopping, a job interview, a breakup). The goal is to find the "human" element in the "non-human."

Step 2: Character Consistency (The LoRA Method)

To keep the monster looking the same, you need a consistent reference. Use a tool like Midjourney with --cref (Character Reference) or train a Flux LoRA on a specific "monster" design. Ensure you define the "invariants": the plaid shirt, the specific scar pattern, and the skin tone.

Step 3: Keyframe Generation

Generate 10-15 high-quality "hero" images. Use prompts that specify the lighting: "cinematic teal and orange," "moody low-key lighting," "overcast bus window light." These will be your anchors for the video generation.

Step 4: Video Generation (Image-to-Video)

Use Runway Gen-3 Alpha or Luma Dream Machine. Upload your keyframes and use "Motion Brush" or specific camera prompts (e.g., "slow dolly zoom," "handheld camera shake") to bring the scenes to life. Keep the motion subtle to maintain the "film" feel.

Step 5: Lip-Sync & Performance

For the speaking parts, use a tool like Hedra or LivePortrait. Upload a clear close-up of your monster and the voiceover audio. This ensures the "acting" feels intentional and the lip-sync is tight.

Step 6: Voiceover & Sound Design

Use ElevenLabs to create a "Deep, melancholic, weary male voice." For the music, find a royalty-free acoustic track or use Udio/Suno to generate a "melancholic French folk song with acoustic guitar."

Step 7: The "Film" Edit

In CapCut or Premiere, add a film grain overlay, slight chromatic aberration, and a letterbox (2.35:1 aspect ratio). This hides "AI artifacts" and makes the video feel like a real movie.

Step 8: The Message Overlay

End with a simple, high-contrast text overlay. Use a clean sans-serif font. The message should be the "moral" of the story, giving the viewer a reason to share it.

Growth Playbook: Distribution & Scaling

3 Opening Hook Lines

"What if the monsters we fear are just as lonely as we are?"
"I used AI to make the saddest Halloween movie you'll ever see."
"Proof that AI can actually make you feel something."

4 Caption Templates

The Emotional Hook: "People don't mean to be cruel. They just don't see you. 🧵 A short film about the scariest thing of all: being alone. [Link to thread]"
The Technical Breakdown: "Character consistency in AI is finally here. I spent 20 hours making sure this 'monster' looked the same in every shot. Here is the result. 👇"
The Holiday Reflection: "Halloween is the one night he feels human. A reminder to check on your 'lonely' friends today. 🎃"
The CTA: "Loneliness is a silent epidemic. If this moved you, send a text to someone you haven't talked to in a while. ❤️"

Hashtag Strategy

Broad: #AI #ShortFilm #Cinematography #Storytelling
Mid-Tier: #RunwayGen3 #AIVideo #Frankenstein #Halloween2025
Niche: #IndieCreator #CharacterConsistency #VisualStorytelling #LonelinessAwareness

Frequently Asked Questions

What tools make it look the most similar?

Runway Gen-3 Alpha for video and Midjourney --cref for character consistency are the current gold standards.

What are the 3 most important words in the prompt?

"Cinematic," "Desaturated," and "Subsurface-scattering" (for realistic skin texture).

Why does the generated face look inconsistent?

Usually due to lack of a strong image reference or too much motion in the prompt; use lower motion settings (3-4).

How can I avoid making it look like AI?

Add film grain, use shallow depth of field (bokeh), and avoid fast, unnatural camera movements.

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

Instagram Reels favors this "high-aesthetic" cinematic style, while TikTok prefers the "behind-the-scenes" process.

How should I properly disclose AI use?

Use the "AI Generated" label provided by the platform and mention the tools used in the first comment or caption.