0:00 / 0:00

Adriana Bubori

@adrianabubori

INSTAGRAM · 2026-02-23Source

128likes

265comments

Remix This

Recreate with Runway Gen 4.5

Make your own AI viral video

Prompt

GLOBAL LOCK:
The presenter is a Caucasian woman in her mid-30s with long, straight blonde/light brown hair. She wears a dark brown sleeveless turtleneck top. The setting is a minimalist room with a large abstract painting in beige and blue tones behind her. She sits at a light-colored wooden table. The lighting is soft, natural, and high-key. The AI-generated model is a Caucasian woman with dark hair, green eyes, and an oval face, wearing an olive green suit. The overall color grade is warm, neutral, and editorial.

[00:00–00:03]
The presenter is in a medium close-up, speaking directly to the camera with expressive hand gestures. Large white bold text overlays: "this is how" then "consistent images" then "with a model". The camera is static.

[00:03–00:05]
Full-screen title card. Background is a solid muted brown. Large white text: "Step 1" with a small icon of a model's face, and "Generate Close-up" below it.

[00:05–00:10]
Back to the presenter in MCU. She holds up a virtual white card showing a high-detail close-up of the AI model's face. Small text labels with lines point to the face: "green eyes", "dark hair", "full lips", "oval-shaped". She is explaining the importance of facial details.

[00:10–00:15]
Full-screen title card. Muted brown background. Text: "Step 2" with an icon of a suit, and "Define Outfit" below it. Transition to a quick shot of the presenter talking with a screenshot of an AI prompt interface showing an olive green suit.

[00:15–00:18]
Full-screen title card. Muted brown background. Text: "Step 3" with a grid icon, and "Lock Identity" below it.

[00:18–00:24]
The presenter is in MCU, talking. A large grid of 9 images overlays the screen, showing the AI model in the olive green suit from various angles and with different facial expressions (smiling, serious, looking away). Text "different angles" and "different facial" overlays the grid.

[00:24–00:28]
Full-screen cinematic shots of the AI model. Shot 1: The model sitting on the floor in the olive suit, looking at the camera. Shot 2: A closer shot of the model sitting, hand on chin. Text "for any" and "you're going to generate" overlays these shots.

[00:28–00:32]
Back to the presenter in MCU. She gestures towards the camera. A black pill-shaped button with the "invideo" logo appears. Finally, a white text box with "comment MODEL" appears at the top. She is giving the final call to action.

NEGATIVE PROMPT:
Visual: blurry faces, inconsistent eye color, distorted limbs, flickering background, low resolution, messy hair, unrealistic skin texture, text watermarks, logos on clothing.
Speech: robotic tone, monotone delivery, background noise, muffled audio, lip-sync mismatch, long awkward pauses, stuttering.

SPEECH PACK:
[00:00-00:03]
Transcript: "This is how to generate consistent images with a model with AI."
TAKE_A: (Energetic, fast-paced) "This is how to generate consistent images with a model with AI!"
TAKE_B: (Authoritative, measured) "This is how... you generate consistent images... with a model... using AI."
TAKE_C: (Friendly, helpful) "Here is exactly how to get consistent images of your AI model."

[00:05-00:10]
Transcript: "First, generate a close-up of your model to capture all the facial details."
TAKE_A: "Step one: generate a close-up of your model to lock in those facial details."
TAKE_B: "First, you need a high-res close-up... to capture every single facial detail."

[00:28-00:32]
Transcript: "And you can do all of this in one single tool, InVideo. Comment MODEL and I'll send you the link."
TAKE_A: "Do it all in one tool: InVideo. Just comment MODEL for the full process!"
TAKE_B: "Everything happens in InVideo. Comment the word MODEL and I'll DM you the link right now."

PROSODY NOTES:
- Emphasis on "consistent" (00:01)
- Pause after "Step 1" (00:03)
- Rising intonation on "Comment MODEL" (00:30)
- Clear, crisp enunciation throughout.

How adrianabubori Made This AI Model Consistency Tutorial Workflow — and How to Recreate It

This case study analyzes a high-performing AI tutorial by creator @adrianabubori, which masterfully simplifies a complex technical problem: AI character consistency. The video uses a "talking head" format set in a minimalist, high-end interior—featuring a large abstract canvas and warm, natural lighting—to establish authority and aesthetic appeal. By breaking down the workflow into three digestible "Steps" (Generate Close-up, Define Outfit, Lock Identity), the creator transforms a daunting technical hurdle into an achievable win for her audience. The visual style is clean and "editorial," mirroring the high-fashion aesthetic of the AI-generated model shown in the examples. This isn't just a tutorial; it's a lead-generation machine that uses a specific "Comment [Keyword]" call-to-action to drive engagement and automate community building. For indie creators, this video serves as a blueprint for "Educational Aesthetic" content that converts viewers into followers and leads.

What You’re Seeing: A Visual Breakdown

The video features a female presenter in a medium close-up, positioned behind a light-colored wooden table. The background is a sophisticated, neutral-toned room with a large piece of abstract art, creating a professional yet accessible "home office" vibe. The lighting is soft and directional, coming from the front-left, which creates gentle shadows and highlights her facial features. The color palette is strictly controlled: beige, brown, and olive green, which creates a cohesive, high-end "editorial" feel throughout the 32-second duration.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:03	Presenter speaking with large white text overlays: "this is how consistent images."	Medium Close-Up (MCU), static.	Warm, high-key, natural.	Hook: Immediate problem-solution statement.
00:03–00:05	Title Card: "Step 1: Generate Close-up."	Graphic overlay on brown background.	Flat, high-contrast text.	Structure: Signposting the learning journey.
00:05–00:10	Presenter holds a virtual "card" showing a model's face. Labels point to "green eyes," "dark hair."	MCU with dynamic overlays.	Consistent with hook.	Demonstration: Showing the "input" required.
00:10–00:15	Title Card: "Step 2: Define Outfit." Followed by a prompt interface screenshot.	Graphic + Screen recording feel.	Neutral, tech-focused.	Process: Demystifying the AI tool interface.
00:15–00:24	Title Card: "Step 3: Lock Identity." Grid of 9 images showing the model in various poses.	Graphic + Multi-image grid.	Editorial, high-fashion.	Proof: Showing the "consistency" promised in the hook.
00:24–00:28	Full-screen cinematic shots of the model in an olive suit.	Full-screen AI generation.	Cinematic, soft shadows.	Aspiration: Showing the high-quality end result.
00:28–00:32	Presenter returns to screen with "Comment MODEL" overlay.	MCU, direct eye contact.	Warm, inviting.	Conversion: Clear CTA to drive engagement.

Why It Went Viral: The "Lead Magnet" Strategy

The Topic: Solving the "AI Consistency" Pain Point

The core of this video's success is its utility. For anyone experimenting with AI image generation (Midjourney, Stable Diffusion, etc.), maintaining the same face across different shots is the "Holy Grail." By addressing this specific, high-intent problem, the creator ensures her content is highly "saveable." Psychologically, it taps into the "Aha!" moment—viewers feel they are getting a professional secret for free. The choice of a high-fashion model as the subject also taps into the "aesthetic" trend on Instagram, making the video visually pleasing even to those not interested in the technical side.

Platform Perspective: Engagement Loops

From a platform signal perspective, this video is a masterclass in Watch Time and Save Rate. The "Step 1, 2, 3" structure creates an open loop that keeps viewers watching until the end to see the final result. However, the real "viral" engine is the CTA: "Comment MODEL." This does three things: 1. It signals to the Instagram algorithm that the video is highly engaging, triggering more reach. 2. It allows the creator to use automation (like ManyChat) to send the full tutorial/link via DM, moving the relationship from public to private. 3. It reduces the "friction" of the viewer having to find a link in the bio, increasing the conversion rate of viewers to leads.

5 Testable Viral Hypotheses

The "Numbered List" Bias: Using "Step 1, 2, 3" reduces cognitive load. Replicate by: Breaking any complex process into exactly three steps.
The "Aesthetic Authority" Effect: The presenter's high-end environment makes the technical advice feel more "premium." Replicate by: Ensuring your background art and lighting match the quality of the output you are teaching.
The "Keyword Trigger" Loop: Asking for a specific comment (e.g., "MODEL") creates a massive spike in engagement. Replicate by: Offering a "free worksheet" or "prompt list" in exchange for a specific comment.
The "Visual Proof" Hook: Showing the final, consistent model grid at 0:18 proves the method works before the video ends. Replicate by: Showing the "After" result early in the "How-to" process.
The "Editorial Palette" Consistency: The presenter's brown top matches the model's olive suit and the beige background. Replicate by: Color-grading your tutorial footage to match your AI-generated examples.

How to Recreate: From 0 to 1

Identify a "How-To" Pain Point: Choose a specific AI problem (e.g., "consistent faces," "realistic hands," "cinematic lighting").
Script for Speed: Write a 30-second script. Hook (0-3s), Step 1 (4-10s), Step 2 (11-18s), Step 3 (19-25s), CTA (26-30s).
Film the "Talking Head": Use a minimalist background. Wear a neutral color. Use soft, front-facing natural light.
Generate Your AI Assets: Use a tool like Midjourney or InVideo to create your "consistent" model. Ensure you have a close-up, an outfit shot, and a grid of various poses.
Create Title Cards: Use a simple, bold font (like Helvetica or Inter) on a solid background that matches your video's color palette.
Edit with "Punch-ins": In CapCut, add slight zooms or "punch-ins" on key phrases to keep the visual energy high.
Overlay the Assets: Use the "Overlay" or "Picture-in-Picture" feature to show the AI images as you talk about them. Use labels (e.g., "green eyes") to highlight details.
Set Up Automation: Use a tool like ManyChat to set a "Comment Trigger" for your chosen keyword so you can automatically DM the "Full Workflow" to viewers.

Growth Playbook: Distribution & Scaling

3 Ready-to-Use Opening Hook Lines

"Stop struggling with AI consistency. This 3-step workflow fixes it."
"How I created a consistent AI brand model in under 5 minutes."
"The secret to locking AI character identity (it’s easier than you think)."

4 Caption Templates

The "Value First" Template:
Stop guessing and start generating. 🚀 This is the exact workflow I use to keep my AI models consistent across every shot.
1️⃣ Generate the base face.
2️⃣ Define the wardrobe.
3️⃣ Lock the identity.
Want the full prompt list? 👇
Comment "MODEL" and I’ll DM it to you for free!

Hashtag Strategy

Broad (Reach): #AI #ArtificialIntelligence #ContentCreation #DigitalMarketing
Mid-Tier (Niche): #AIArt #MidjourneyTutorial #AIWorkflow #IndieCreator
Long-Tail (Specific): #ConsistentAI #AIModelGeneration #InVideoAI #AITipsAndTricks

Frequently Asked Questions

What tools make it look the most similar?

Midjourney for the images and InVideo or CapCut for the video editing and overlays.

What are the 3 most important words in the prompt?

"Consistent," "Identity," and "Reference."

Why does the generated face look inconsistent?

Usually because the initial "base" image doesn't have enough distinct facial features for the AI to track.

How can I avoid making it look like AI?

Use "photorealistic" and "raw" style tags, and ensure the lighting in the AI image matches the "real" world.

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

Instagram, as the "aesthetic" and "saveable tutorial" format performs exceptionally well there.