0:00 / 0:00

Adriana Bubori

@adrianabubori

INSTAGRAM · 2026-03-05Source

324likes

549comments

Remix This

Recreate with Runway Gen 4.5

Make your own AI viral video

Prompt

GLOBAL LOCK: Two primary female subjects. Subject A: Black woman, mid-20s, voluminous natural curly hair, warm skin tone, wearing a beige/cream oversized blazer and matching trousers. Subject B: Caucasian woman with red curly hair and prominent freckles, mid-20s, wearing a dark chocolate brown blazer and trousers. Environment: Minimalist studio with solid color backgrounds (beige, soft pink). Lighting: High-end cinematic editorial lighting, soft-box style, gentle shadows, vibrant color palette (orange, red, pink). Camera: Professional 35mm lens feel, shallow depth of field, dynamic movement. Speech: Energetic female voiceover, crisp studio recording, high lip-sync strictness for on-camera moments.

[00:00–00:02]
Subject A is in a medium close-up, leaning her arms over the curved edge of a large orange slide. She is looking directly at the camera, speaking with an engaging expression. Background is a clean, solid beige. Lighting is bright and even.
Speech: "So I heard you want to generate"
Lip-sync: High.

[00:02–00:04]
Wide shot, high angle. A long, orange spiral slide winds down against a beige wall. A model in a beige suit slides down the curves. The camera pans slightly to follow the movement.
Speech: "campaigns for your brand that are creative and fast"

[00:04–00:06]
Subject A is sitting on a bright red seesaw, medium shot. She is holding the handles and laughing. Background is a soft pastel pink.
Speech: "that allow you to"

[00:06–00:07]
Hard cut to the exact same framing and seesaw, but Subject A is replaced by Subject B (redhead). She is in the same pose, laughing. This is a seamless "model swap."
Speech: "change the colors, swap models"

[00:07–00:09]
Subject B is standing in a full shot against the pink background. She holds a red bowl of dog food. A long-haired brown dachshund stands at her feet, looking up. She bends down slightly.
Speech: "and even make them hold or wear"

[00:09–00:10]
Close-up of the dachshund eating from the red bowl on the floor. The dog's fur texture is detailed.
Speech: "your products"

[00:10–00:13]
A dynamic grid animation showing multiple rectangular frames of both models in various poses and outfits (beige and brown suits). The frames tilt and move slightly.
Speech: "for your socials, your ads, and even your website"

[00:13–00:15]
Subject A is sliding down a red slide, medium shot, smiling. The camera tracks her movement. Beige background.
Speech: "Sounds fun, right?"

[00:15–00:16]
Subject B is walking across a red and white striped crosswalk painted on the floor. Side profile, medium shot. Pink background.
Speech: "The thing is"

[00:16–00:18]
Wide shot of Subject A sliding down the red slide, arms out, laughing.
Speech: "smart brands build campaigns"

[00:18–00:19]
Subject B is crouching on the red and white crosswalk, looking at the camera and laughing.
Speech: "that are worth remembering"

[00:19–00:21]
Extreme close-up profile of Subject B smiling and looking off-camera. Focus on her freckles and red hair texture.
Speech: "And thanks to AI, that's now easier than ever"

[00:21–00:25]
Subject B is in a direct-to-camera close-up, speaking directly to the viewer with a friendly, professional expression. Dark brown blazer visible.
Speech: "Comment AI and we'll send you the link to learn how"
Lip-sync: High.

NEGATIVE PROMPT: Uncanny valley, distorted facial features, inconsistent hair texture between shots, blurry skin, robotic body movement, flickering lighting, text artifacts, messy background, low resolution, mismatched lip-sync, harsh sibilance in audio, background noise.

SPEECH PACK:
[00:00-00:02] "So I heard you want to generate"
TAKE_A: (Curious, upbeat) So... I heard you want to generate...
TAKE_B: (Direct, professional) So I heard you want to generate...
TAKE_C: (Friendly, conspiratorial) So, I heard *you* want to generate...

[00:21-00:25] "Comment AI and we'll send you the link to learn how"
TAKE_A: (Action-oriented) Comment AI, and we'll send you the link to learn how!
TAKE_B: (Helpful) Just comment AI, and we'll send over the link.
TAKE_C: (Energetic) Comment AI now, and we'll send you the link to learn how!

How Adriana Bubori Made This Brand Campaign Model Swap AI Video — and How to Recreate It

This case study analyzes a high-energy, cinematic AI brand campaign created by @adrianabubori. The video showcases the power of generative AI to create professional-grade marketing content that is both visually stunning and highly flexible. Featuring a vibrant palette of bold oranges, reds, and soft pinks, the video demonstrates "model swapping" and product integration within a minimalist, editorial studio setting. It perfectly captures the "AI Aesthetic 2.0"—moving away from uncanny valley weirdness toward high-fashion, high-production value content that looks like it belongs in a luxury magazine or a premium TV spot.

What You’re Seeing: Visual & Audio Analysis

The video is a masterclass in dynamic pacing and color coordination. We see two primary subjects: a Black woman with voluminous natural curls in a beige power suit, and a redheaded woman with freckles in a chocolate brown blazer. The scenes are set against clean, monochromatic backgrounds, utilizing large-scale props like a spiral orange slide and a red seesaw to create a sense of "adult playfulness."

The lighting is soft-box studio lighting, providing even coverage with gentle highlight roll-offs on the skin, emphasizing texture and realism. The camera work is active, featuring tracking shots that follow the models down slides and quick "punch-in" edits that sync with the upbeat, rhythmic background track. The voiceover is crisp and professional, delivered with an encouraging, "insider" tone that positions AI as a creative superpower rather than just a tool.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Color	Viewer Intent
00:00–00:02	Model A leaning over an orange slide, speaking.	Medium Close-Up (MCU)	Warm beige/orange tones.	Hook: Direct address to the viewer.
00:02–00:04	Wide shot of a model sliding down a spiral orange slide.	Wide Shot (WS), High Angle	Minimalist beige background.	Visual Interest: Dynamic movement.
00:04–00:07	Model A on a red seesaw, then "swaps" to Model B.	MCU, Static	Soft pink background, red prop.	Feature Demo: Showcasing AI model swapping.
00:07–00:10	Model B feeding a dachshund; close-up of dog eating.	Full Shot to Close-Up	Pink/Red palette.	Emotional Value: Cute animal + product use.
00:10–00:13	Fast-paced grid collage of various shots.	Graphic Overlay / Grid	Multi-color high energy.	Social Proof: Showing versatility of output.
00:13–00:18	Models sliding and walking across striped crosswalks.	Tracking & Wide Shots	Red/White high contrast.	Reinforce Persona: High-end fashion vibe.
00:18–00:25	Model B laughing, then direct-to-camera CTA.	Close-Up (CU)	Warm, freckled skin detail.	Conversion: Clear call to action.

Why It Went Viral: The Mechanics of Engagement

The "Problem-Solution" Hook

The video targets a massive pain point for small business owners and indie creators: the high cost and slow speed of traditional content production. By starting with "I heard you want to generate campaigns... that are creative and fast," it immediately filters for an audience looking for efficiency. It doesn't just show "cool AI"; it shows a business solution.

Visual Satisfaction & The "Model Swap" Magic

Psychologically, the human brain loves patterns and smooth transitions. The seamless swap between the two models on the seesaw (around 0:06) acts as a "pattern interrupt." It’s a subtle demonstration of technical mastery that makes the viewer pause and wonder, "How did they keep the lighting and prop exactly the same?" This increases the "re-watch" factor as users try to spot the seam.

Platform Signals & Engagement Loops

From a platform perspective, this video is engineered for the Instagram/TikTok algorithm. The fast cuts (averaging 1.5 seconds per shot) keep watch time high. More importantly, the "Comment AI" CTA is a classic engagement hack. By forcing users to comment a specific keyword to get a link, the creator triggers a flood of engagement signals, telling the algorithm the content is highly relevant, which pushes it to a wider audience.

5 Testable Viral Hypotheses

The "Keyword Trigger" Effect: If you ask for a specific keyword in the comments to receive a "secret" link, engagement will increase by 3x compared to a standard "link in bio" CTA.
The "Aesthetic Authority" Theory: High-saturation, color-coordinated visuals (e.g., all orange/red) establish instant brand authority, making viewers more likely to trust the "tutorial" or "service" being offered.
The "Animal Integration" Boost: Including a high-quality AI-generated animal (like the dachshund) increases "saves" because it demonstrates a level of AI control that is currently difficult to achieve.

The "Model Swap" Hook:

The "Text-on-Screen" Retention: Using large, bold, centered text that mirrors the voiceover helps viewers consume the content even with sound off, increasing overall completion rates.

How to Recreate: From 0 to 1

Step 1: Define Your Color Story

Pick 2-3 dominant colors. In this video, it's Orange, Red, and Pink. Consistency in background and props is key to making AI look "expensive."

Step 2: Character Design (Consistency)

Create a detailed character sheet for your models. Use tools like Midjourney or Flux to generate 10+ reference images of the same person in different poses. Prompt Tip: "Editorial fashion photography, [Ethnicity] woman, [Hair Style], wearing [Specific Outfit], minimalist studio background, soft-box lighting, 8k resolution."

Step 3: Generate the "Base" Scenes

Use a video AI tool (like Kling, Luma Dream Machine, or Runway Gen-3) to create the motion. Start with the "Model Swap" scene. Generate the same action (e.g., sitting on a seesaw) twice, changing only the subject description in the prompt while keeping the environment description identical.

Step 4: Master the "Product Integration"

To show a model holding a product, use "Image-to-Video" features. Upload a high-quality photo of your product and use a prompt like: "Model's hands reaching down to pick up the [Product Name], cinematic movement, focus on product."

Step 5: Voiceover & Scripting

Use ElevenLabs to generate a high-quality, energetic voiceover. Choose a "Professional" or "Narrator" voice profile. Ensure the script follows the Hook → Value → CTA structure seen in the video.

Step 6: Editing for Rhythm

In CapCut or Premiere, cut your clips to the beat of your music. No shot should last longer than 2 seconds unless it's the final CTA. Add bold, white sans-serif text overlays (e.g., Montserrat or Helvetica Bold) in the center of the frame.

Step 7: The Engagement Automation

Use a tool like ManyChat to automate the "Comment AI" response. Set it up so that whenever someone comments "AI," they automatically receive a DM with your link.

Growth Playbook: Distribution & Scaling

3 Ready-to-Use Opening Hooks

"Stop spending $5k on brand shoots. Do this instead..."
"I found the secret to swapping models in your ads with one click."
"Want to see how the biggest brands are using AI right now?"

4 Caption Templates

The "Efficiency" Angle: "Content creation used to take weeks. Now it takes hours. ⚡️ Here is how we built this entire campaign using only AI. Want the tool list? Comment 'TOOLS' below!"
The "Aesthetic" Angle: "Visuals are the new currency. 💎 We’re pushing the limits of AI consistency to create high-end editorial looks for a fraction of the cost. Which shot was your favorite? 1 or 2?"
The "Tutorial" Angle: "Step-by-step: How to swap models in the same scene without losing quality. 🧵 (Save this for your next project!) Comment 'LEARN' for the full breakdown."
The "Direct" Angle: "AI isn't coming for your job, it's coming for your workflow. 🚀 Build campaigns that are worth remembering. Comment 'AI' to see how we do it."

Hashtag Strategy

Broad (Reach): #AI #ContentCreation #MarketingTips #DigitalArt
Mid-Tier (Targeted): #AIVideo #GenerativeAI #BrandStrategy #IndieCreator
Niche (High Intent): #KlingAI #RunwayGen3 #AIForBusiness #ModelConsistency

Frequently Asked Questions

What tools make it look the most similar?

Use Flux.1 for the base images and Kling AI or Luma for the most realistic human motion.

What are the 3 most important words in the prompt?

"Editorial," "Minimalist," and "Cinematic Lighting" are essential for this specific high-end look.

Why does the generated face look inconsistent?

You likely aren't using a "Character Reference" (Cref) or a consistent seed; ensure your base character description is identical across all prompts.

How can I avoid making it look like AI?

Focus on "soft lighting" and "film grain" in your prompts to avoid the overly smooth, plastic look of default AI generations.

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

Instagram currently favors high-aesthetic "Aesthetic AI," while TikTok favors "How-to" and "Behind the scenes" versions of this content.