Create Viral POV Found Footage AI Horror Videos (Step-by-Step Guide)

🛠️ Tools & Resources Used

ChatGPT (LLM / Scripting) - Generates viral concept ideas and tailored two-part prompt sequences.
Google Flow (AI Video Generator) - Renders high-fidelity vertical POV videos with start-frame image consistency.
CapCut (Video Editor) - Stitches clips together and layers immersive suspense sound design.

⏱️ Quick Workflow Summary

Step 1: Run the Master Prompt in ChatGPT to select an exploration concept and get a two-part prompt sequence.
Step 2: Generate Clip 1 in Google Flow using Omniflash at 9:16 aspect ratio.
Step 3: Extract the final frame of Clip 1 and use it as an image-to-video reference for Clip 2.
Step 4: Assemble both clips in CapCut and add layered Foley and creature sound design.

📝 Step-by-Step Tutorial

Step 1: Brainstorm Concepts and Generate Prompts with ChatGPT [Timestamp: 00:34]

Input the Master Prompt into ChatGPT to generate 10 viral exploration ideas combining a distinct location, a discovered treasure, and a guarding entity. Choose your favorite concept, and let the model provide the tailored two-part prompt sequence (0–10s approach and 10–20s creature reveal).

Prompt / Settings:

PROMPT / SETTINGS

You are an expert AI video prompt director specializing in ultra-realistic first-person POV found-footage exploration videos with a thriller/creature-feature climax.

### Style Signature:
- Perspective: Raw first-person POV shot on a mobile phone or GoPro.
- Lighting: Pitch-black environment cut only by a handheld flashlight beam.
- Camera: Handheld wobble, forward walking, dust motes caught in the light.
- Tension: Environmental clues of danger leading to an encounter with an entity or creature.

---

### Step 1: Brainstorming Concepts
Provide 10 viral-worthy hidden exploration concepts across diverse locations.
For each concept include:
- Title & Location
- The Discovery: What treasure/artifact is found.
- The Danger: What lurking entity, monster, or creature is guarding it.

Wait for me to choose one before proceeding.

---

### Step 2: Two-Part Video Prompt Generation
Once selected, output a 2-part video sequence to build a seamless 20-second video via image-to-video continuation:

#### Workflow Steps:
1. Generate video using **Prompt 1** (0–10s).
2. Extract the exact last frame as an image.
3. Upload that frame as the start image for **Prompt 2** (10–20s).
4. Generate the second clip and stitch together.

---

#### Video Prompts:

**Prompt 1 (0–10s: The Approach & Clues):**
First-person POV, handheld flashlight beam cutting through darkness in [location]. Camera moves forward past warning signs like claw marks, crushed gear, and scattered artifacts. Subtle camera shake, dust particles in the air, phone camera realism.

**Prompt 2 (10–20s: The Treasure & The Threat):**
Continuation from reference image. Camera steps closer to the central treasure pile. The flashlight pans up from the gold/artifacts to reveal [the sleeping/lurking creature] in the dark shadows behind it. Sudden slight camera tremor and dramatic focal shift.

#### Suggested Audio Prompts (for sound design):
- Ambient cavern drips, tense breathing, slow gravel footsteps.
- Low, resonant creature growl or subtle shifting movement in the shadows.

Step 2: Render Clip 1 in Google Flow [Timestamp: 01:03]

Open Google Flow and choose the Omniflash model set to a 9:16 vertical ratio. Enter the first prompt (0–10s Approach) generated by ChatGPT and render your initial exploration footage showing the flashlight cutting through darkness.

Google Flow Omniflash rendering the initial POV exploration clip

Prompt / Settings:

PROMPT / SETTINGS

Model: Omniflash
Aspect Ratio: 9:16 Vertical
Prompt: First-person POV, handheld flashlight beam cutting through darkness in [location]. Camera moves forward past warning signs like claw marks, crushed gear, and scattered artifacts. Subtle camera shake, dust particles in the air, phone camera realism.

Step 3: Extract the Last Frame and Generate Clip 2 [Timestamp: 01:19]

Scrub to the exact last frame of the rendered first video and export it as a high-resolution image. Set this extracted frame as the start image reference in Google Flow, then render the second prompt (10–20s Reveal). This locks in lighting, color temperature, and environmental geometry across the seam.

Creature reveal continuation maintaining lighting and environment consistency

Prompt / Settings:

PROMPT / SETTINGS

Input Reference: Last frame image from Clip 1 (Start Frame)
Prompt: Continuation from reference image. Camera steps closer to the central treasure pile. The flashlight pans up from the gold/artifacts to reveal [the sleeping/lurking creature] in the dark shadows behind it. Sudden slight camera tremor and dramatic focal shift.

Step 4: Stitch and Layer Audio in CapCut [Timestamp: 01:54]

Import both video clips into CapCut and align them on the timeline. Build out the soundscape by adding gravel footsteps, ambient cave drips, and nervous breathing during the approach, followed by a heavy creature growl right as the beam illuminates the monster.

💡 Key Takeaways & Pro Tips

Never Generate Independent Clips: Avoid generating the approach and the creature reveal separately without an image reference; doing so will cause noticeable morphing and inconsistencies.
Audio Hooks Drive Retention: Raw AI video alone will not retain viewers—precise sound cues timed to the flashlight motion are critical to building tension.
Atmospheric Keywords Matter: Include terms like 'dust motes', 'handheld wobble', and 'phone camera realism' to bypass overly clean CGI appearances.

📌 Original Source & Attribution

Source Video: This Untapped Faceless AI Niche is Blowing Up Right Now (Free Workflow)
Original URL: https://www.youtube.com/watch?v=5OaZ5RdMr0Y
Credit: FlowState AI