0:00 / 0:00

▶

Ashmitha Mohan

@cyborggirll

INSTAGRAM · 2026-02-28Source

359likes

9comments

Remix This

Recreate with Runway Gen 4.5

Make your own AI viral video

Prompt

GLOBAL LOCK:
Subject: 20-year-old South Asian woman, warm skin tone, dark hair tied in a sleek ponytail, black sleeveless tank top, minimalist aesthetic.
Environment: Consistent subject identity across all shots, but background transitions from a neutral indoor office to various cinematic outdoor locations.
Lighting: Transitions from soft indoor ambient light to environment-specific lighting (golden hour, harsh sun, neon).
Camera: Eye-level MCU, mostly static with subtle organic movement.
Speech: Warm, energetic female voice, clear articulation, direct-to-camera address.

[00:00–00:03]
Subject: Split-screen view. Top half shows the woman in a neutral indoor office. Bottom half shows her in a misty, sun-drenched forest with ferns.
Action: She looks up and to the right, then turns her head slowly.
Camera: Static split-screen, MCU.
Lighting: Top is flat indoor; Bottom is cinematic golden hour with light rays (god rays) through trees.
Grade: Bottom has a warm, earthy, high-contrast cinematic grade.
Motion: Subtle hair movement, drifting mist in the background.

[00:04–00:07]
Subject: The woman holding a small black wireless microphone close to her mouth.
Action: Speaking directly to the camera, gesturing with her free hand.
Environment: Neutral indoor background with a plant and a modern sculpture.
Camera: MCU, static.
Lighting: Soft natural light from the side.
Speech: "You can take shots without any heavy editing skills, and let me show you how."
Sync: High lip-sync strictness; cut lands on "how."

[00:08–00:14]
Subject: The woman seen through a smartphone camera UI.
Action: She holds up her own phone as if taking a selfie, looking around "silly" at the empty room.
Environment: Indoor office.
Camera: Handheld feel, looking at a phone screen.
Lighting: Standard indoor.
Speech: "Step 1: Record a video of yourself acting like you are in your desired shot. This might look a little silly."
Sync: VO over action.

[00:15–00:19]
Visual: Screen recording of the "Banana 2" web interface. A screenshot of the woman is uploaded.
Action: Cursor clicks "Generate" after a prompt is typed: "Replace indoor setting with a realistic outdoor scene... misty forest... cinematic."
Lighting: UI dark mode.

[00:20–00:28]
Visual: Screen recording of the "Kling" AI interface. The original video and the generated forest image are uploaded.
Action: Cursor clicks "Generate." The result shows the woman seamlessly integrated into the forest.
Speech: "Step 2: Now head to Kling, go to video generation option, add your video and the reference picture. Prompt this, hit generate, and wallah!"
Sync: VO over UI action.

[00:29–00:36]
Visual: Screen recording of ChatGPT.
Action: Typing a request for a "similar prompt for visiting a lively city." ChatGPT generates a new prompt.
Subject: Inset of the woman in the Times Square background, holding her phone.
Speech: "I will leave the prompt in the comments section. And if you want to change the scene, take this prompt to ChatGPT and describe what you want and you will get a brand new prompt."
Sync: VO over UI and inset video.

[00:37–00:40]
Subject: The woman speaking to the camera, holding the mic.
Action: Smiling, pointing to the screen. Text "the CYBORG girl" appears in neon pink.
Environment: Indoor office.
Speech: "Ready? And for cool AI hacks as such, follow the Cyborg Girl for more."
Sync: High lip-sync strictness.

NEGATIVE PROMPT:
Visual: Blurred face, inconsistent hair texture, extra fingers, warped background geometry, flickering lighting, low resolution, watermarks, distorted limbs, unnatural skin smoothing.
Speech: Robotic tone, muffled audio, background noise, lip-sync delay, harsh "S" sounds, inconsistent volume.

SPEECH PACK:
[00:04–00:07]
Transcript: "You can take shots without any heavy editing skills, and let me show you how."
TAKE_A: Warm, encouraging, medium pace.
TAKE_B: Fast, energetic, "hacker" vibe.
TAKE_C: Slow, deliberate, authoritative.

[00:08–00:14]
Transcript: "Step 1: Record a video of yourself acting like you are in your desired shot. This might look a little silly."
TAKE_A: Playful, slightly laughing on "silly."
TAKE_B: Instructional, clear steps.

[00:20–00:28]
Transcript: "Step 2: Now head to Kling, go to video generation option, add your video and the reference picture. Prompt this, hit generate, and wallah!"
TAKE_A: Exciting, building to the "wallah" reveal.
TAKE_B: Direct, tutorial-style.

[00:37–00:40]
Transcript: "Ready? And for cool AI hacks as such, follow the Cyborg Girl for more."
TAKE_A: Friendly, high energy, strong CTA.
TAKE_B: Professional, calm.

How cyborggirll Made This AI Background Replacement Kling — and How to Recreate It

This tutorial by creator @cyborggirll is a masterclass in cinematic AI background replacement. By leveraging the power of "Video-to-Video" AI tools like Kling and image generators like Banana 2, she demonstrates how an indie creator can transform a mundane indoor setting into high-production environments like a misty forest, a vast desert, or the neon-lit streets of Times Square. The video uses a high-contrast "Before vs. After" hook that immediately captures attention, followed by a clear, actionable workflow that demystifies complex AI processes for a non-technical audience. The aesthetic is clean, professional, and highly aspirational, making it a perfect example of how AI can bridge the gap between low-budget setups and premium visual storytelling.

What You’re Seeing: Detailed Visual Analysis

The video features a young South Asian woman with dark hair tied in a sleek ponytail, wearing a simple black sleeveless top. This minimalist wardrobe choice is strategic; it provides clear silhouettes for the AI to track and modify without complex fabric physics issues. The "Before" shots are filmed in a standard home office with a plant and a bookshelf, while the "After" shots showcase the AI's ability to maintain subject consistency while completely overhauling the environment, lighting, and color grade.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Color	Viewer Intent
00:00–00:03	Rapid-fire split-screen "Before/After" comparisons (Forest, Desert, Times Square).	Static split-screen; high-speed cuts.	Warm indoor vs. cinematic outdoor grades.	The Hook: Instant proof of value and visual "wow" factor.
00:04–00:07	Creator speaks directly to the camera holding a small microphone.	Medium Close-Up (MCU); eye-level.	Natural, soft window lighting; neutral tones.	Establish Authority: Introduce the tutorial concept.
00:08–00:14	Screen recording of the "silly" base video being filmed.	POV / Screen recording.	Standard smartphone camera look.	Relatability: Show the "unpolished" starting point.
00:15–00:19	Interface of Banana 2 generating the background image.	UI walkthrough.	Dark mode interface; high contrast.	Instructional: Show the first tool in the stack.
00:20–00:28	Interface of Kling AI processing the video-to-video generation.	UI walkthrough.	Software UI.	Instructional: Show the core "magic" step.
00:29–00:36	Using ChatGPT to generate and refine prompts.	UI walkthrough.	Clean text-based UI.	Value Add: Solve the "what do I type?" friction point.
00:37–00:40	Creator concludes with a CTA and "The Cyborg Girl" logo.	MCU with text overlays.	Soft, flattering indoor light.	Retention/Growth: Brand reinforcement and follow prompt.

Why It Went Viral: The Breakdown

This video taps into the "AI Democratization" trend. It takes a high-end visual effect—something that previously required a green screen, a VFX team, and hours of rotoscoping—and packages it as a simple two-step process for anyone with a smartphone. The "Before/After" format is a psychological goldmine; it triggers curiosity (how did she do that?) and provides immediate gratification. By showing herself acting "silly" in the base video, the creator lowers the barrier to entry, making the viewer feel like they could easily replicate the results.

From a platform perspective, the video excels in Watch Time and Saveability. The fast-paced editing and rapid transitions in the first 3 seconds ensure users don't scroll past. The tutorial nature of the content makes it highly "saveable," as users will want to refer back to the tool names (Banana 2, Kling) and the specific prompt structures shown. The use of trending AI tools like Kling, which is currently a hot topic in the creative tech space, helps the video ride the wave of platform-wide interest in AI content.

5 Testable Viral Hypotheses

The Contrast Hook: If you show a 10/10 cinematic result next to a 2/10 "behind the scenes" shot in the first 2 seconds, watch time will increase by 40% because of the "transformation" curiosity.
The Tool-Stack Reveal: Explicitly naming 2-3 specific tools (Kling, Banana 2) increases saves because viewers treat the video as a resource bookmark.
The "Silly" Factor: Showing the unpolished, slightly embarrassing process of filming the base video increases comments and relatability, as it breaks the "perfect creator" facade.
Prompt Transparency: Showing the exact text used in ChatGPT reduces the "blank page" fear for viewers, leading to higher shares among aspiring creators.
The Loop Effect: Ending the video with a final cinematic result that mirrors the opening hook creates a seamless loop, encouraging multiple views.

How to Recreate: Step-by-Step Guide

Positioning: This content suits "AI Educator," "Digital Artist," or "Tech-Savvy Lifestyle" accounts. Focus on the "magic" of the transformation.
The Base Video: Record yourself in a well-lit, simple indoor environment. Wear a solid-colored outfit (like the black tank top in the video) to help the AI distinguish your body from the background.
The Action: Act as if you are in the target environment. If it's a forest, look up at the "trees." If it's a city, look around at the "lights." This ensures your physical movement matches the final AI environment.
Screenshot Capture: Take a high-quality screenshot of the very first frame of your video. This will be your "Reference Image."
Background Generation (Banana 2 / Midjourney): Upload your screenshot to an AI image generator. Use a prompt like: "Replace indoor setting with a realistic misty forest, natural daylight, cinematic, ultra-realistic." Keep the subject's pose identical.
Video Animation (Kling / Runway Gen-2): Upload your original "silly" video and your new "cinematic" image to a Video-to-Video AI tool. Use the image as the "Character/Environment Reference" and the video as the "Motion Reference."
Prompt Refinement: Use ChatGPT to generate variations. Ask: "I have a prompt for a forest scene, give me a similar prompt for a neon Times Square setting at night."
Editing: Use a split-screen effect for the intro. Add fast-paced text overlays (e.g., "Step 1," "Step 2") and a clear, energetic voiceover.
Publishing: Use a high-contrast cover image showing the "Before" and "After" side-by-side.

Growth Playbook: Distribution & Scaling

Opening Hook Lines

"Stop paying for expensive travel—just use this AI trick."
"How I turned my bedroom into Times Square in 30 seconds."
"The secret tool cinematic creators are hiding from you."

Caption Templates

Option 1: The "Value First" Approach
How to travel the world without leaving your room. 🌍✨
1️⃣ Record a base video.
2️⃣ Use Banana 2 for the background.
3️⃣ Animate with Kling AI.
Which location should I try next? Comment below! 👇
#AIvideo #KlingAI #ContentCreatorTips

Hashtag Strategy

Broad (Reach): #AI #VideoEditing #ContentCreation #TechTrends
Mid-Tier (Niche): #AIArt #KlingAI #GenerativeVideo #IndieCreator
Long-Tail (Specific): #AITutorial #BackgroundReplacement #CinematicAI #Banana2

Frequently Asked Questions

What tools make it look the most similar?

Kling AI for video-to-video and Banana 2 or Midjourney for the initial background image replacement.

What are the 3 most important words in the prompt?

"Cinematic," "Ultra-realistic," and "Natural lighting" are key for high-end results.

Why does the generated face look inconsistent?

Ensure your base video has clear lighting and you use a high-strength "Character Reference" setting in the AI tool.

How can I avoid making it look like AI?

Keep the motion subtle and ensure the lighting on your face in the base video matches the lighting in your prompt.

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

Instagram prefers the high-aesthetic "Before/After" look, while TikTok favors the "Step-by-Step" tutorial breakdown.

How should I properly disclose AI use?

Use the platform's "AI-generated" label and mention the tools in your caption to build trust with your audience.