Meet Dreamina Seedance 2.5 with Precise Segment Editing.
Try Now!

I Made an Analog Horror Broadcast With AI—and Hid the Story in the Glitches

I built an original 1987 training broadcast in Pippit and used repeated shots, timestamps, object counts, and restrained VHS damage to hide the story.

Pippit AI film maker webpage with a prompt box and Generate button on a pink-blue gradient background
Pippit
Pippit
Sep 2, 2026

The fastest way to flatten analog horror is to cover every frame with random static. In this AI film test, I treated video damage as punctuation. The story sits underneath it: a fictional appliance-company training tape repeats one break-room instruction three times, while a clock and a mug quietly stop agreeing with the narration.

Pippit AI film maker webpage with a prompt box and Generate button on a pink-blue gradient background

Figure 1. The AI Film page where I began the fictional training-tape project.

I Wrote the Prop List Before the Plot

The set contains one coffee maker, three mugs, one clock, and one locked door. That inventory is deliberately boring. Repetition teaches you where everything belongs, so a mug beside the door becomes a story event. An analog horror AI concept benefits from countable props because the viewer can verify the change without dialogue.

The Full Production Prompt

Create a 16:9, 20-second original analog-horror training broadcast from a fictional 1987 appliance company called NORTHLINE HOME SYSTEMS. Do not imitate any existing horror series, broadcaster, logo, character, emergency alert, or real company. Open on a blue training card reading 'NORTHLINE HOME SYSTEMS - NIGHT CLEANING PROCEDURE - TAPE 04.' Cut to a beige employee break room containing one chrome drip coffee maker, three white mugs, a wall clock, and a locked supply door. A calm neutral narrator says: 'At the end of your shift, unplug the machine and return all cups to the marked shelf.' Repeat the same instructional shot three times. On the second pass, the wall clock moves backward by one minute. On the third pass, one mug appears beside the locked door while all other objects stay fixed. Add mild VHS scan lines, one tracking roll, stable readable title cards, low tape hiss, and a final card: 'ITEM COUNT DOES NOT MATCH.' No violence, blood, threats, flashing emergency graphics, copyrighted music, or claims that the footage is real.

A Three-Pass Structure Carried the Mystery

Pass
Instructional Layer
Hidden Change
First
Unplug the machine and return the cups.
Nothing is wrong; the room establishes its baseline.
Second
The same procedure repeats.
The wall clock moves backward by one minute.
Third
The same procedure repeats again.
One mug appears beside the locked door.
Final card
The tape stops teaching.
ITEM COUNT DOES NOT MATCH reframes the repeated scene.
Break room with coffee maker, cups on a shelf, a wall clock, and a closed door

Figure 2. The baseline, disrupted repeat, and final evidence from this article's standalone broadcast.

The Glitch Budget Kept the Evidence Visible

I requested mild scan lines, one tracking roll, and low tape hiss. Those effects mark the medium without obscuring the clock or mug. If every cut tears, the viewer stops distinguishing a deliberate clue from a rendering artifact. For a VHS horror video, restraint is not cosmetic; it protects the plot.

Why the Fiction Label Matters

I used an invented company, original title card, and no emergency-alert language. The broadcast is presented as a designed fiction, not leaked footage or a real warning. That boundary keeps the creative device clear while still letting the archival texture do its work. In an AI film workflow, this fiction boundary should be explicit before visual styling begins.

Cleaner Text-and-Prop Revision

Create the same fictional NORTHLINE training room in a locked 16:9 composition. Use only four readable text elements: company name, tape number, one procedure sentence, and final card ITEM COUNT DOES NOT MATCH. Repeat the identical room three times. First repeat: no change. Second: clock moves backward one minute. Third: exactly one mug moves beside the door. Apply light scan lines after the evidence is readable; use one tracking roll between repeats.

Blue title card in a bare room reads, "NORTHLINE HOME SYSTEMS NIGHT CLEANING PROCEDURE - TAPE 04."

Figure 3. Pippit's project record showing the precise fictional broadcast request.

FAQs About Analog-Horror Evidence

Q1. Do VHS Effects Automatically Create Analog Horror?

No. The format supplies texture; repetition and altered evidence supply the story.

Q2. How Much Generated Text Should I Trust Inside a Scene?

Use as little as possible and inspect every AI film frame. Plot-critical cards are safer when added or corrected in the editor after the visual render.

Q3. Can I Imitate a Familiar Broadcast or Horror Series?

Build an original company, layout, terminology, and sound pattern instead. The genre is reusable; another work's identity is not.

Q4. How Much Noise Can a Clue Survive?

Reduce the effect until the object, number, or movement remains identifiable on a phone screen. Texture should delay recognition slightly, not erase the evidence.

Q5. Should Generated Title Cards Stay in the Final Cut?

Only when every important character is stable and readable. Replace unstable lettering in the editor instead of asking viewers to decode accidental text.

The Glitches Carried the Clues

I would keep the repeated room, the backward clock, and the displaced mug, then correct any unstable letters before export. The strongest AI horror film is not the noisiest one. An AI film rewards a second viewing when the glitch points toward evidence instead of covering it.

Continue the workflow: Seedance or AI movie maker. Use them when the archive concept needs a different production scale.

Hot and trending