0:00 / 0:00

The best video models of 2025 👏 Comment “AI” for a link #aivideo #freepik #videogeneration #generativeai #creativeai

Rourke Sefton-Minns

@rourke · creator

INSTAGRAM · 2025-12-17Source

1.7Klikes

558comments

Remix This

Recreate with Runway Gen 4.5

Make your own AI viral video

Prompt

GLOBAL LOCK: The video features a consistent talking-head subject, a Caucasian male with a brown beard, wearing a green and white "Vans" trucker hat and a white t-shirt. He is positioned in a circular overlay with a soft white glow. The background consists of a series of high-end cinematic AI-generated video clips. The overall style is a tech-review/tutorial hybrid. Lighting for the creator is warm and soft; background clips vary from high-key fashion to moody cinematic drama. Color grade is vibrant with high contrast. Speech is energetic, clear, and informative.

[00:00–00:02]
Visual: A 3x3 grid of AI video thumbnails. Each thumbnail has a label: "Kling 2.6", "Runway Gen 4", "Pixverse 5.5", "Sora", "Hailuo 2.3", "Veo 3.1", "Seadance 1.0". The camera zooms slightly into the center.
Subject: Creator in a circular overlay in the center.
Speech: "There's a lot of great AI video models out there."
Sync: Cut to next shot on "out there."

[00:02–00:05]
Visual: Background shows a hyper-realistic close-up of a woman's face with yellow eyeliner and freckles (Seadance 1.0). A UI card appears on the left with "Seadance 1.0" and 4 rating dots for Cost, Speed, and Quality.
Subject: Creator in circular overlay at the bottom.
Speech: "But which one should you be spending your hard-earned money on?"

[00:05–00:08]
Visual: Background shows a man in a grey jacket walking away in a misty, black-and-white mountain landscape (Kling 2.6). UI card updates to "Kling 2.6" with different ratings.
Subject: Creator points up towards the card.
Speech: "Which one is the most cost-effective?"

[00:08–00:10]
Visual: Background shows a woman in a pink suit walking between two black horses on a white salt flat (Runway Gen 4). UI card updates to "Runway Gen 4".
Subject: Creator gives a thumbs up.
Speech: "And what's going to give you the best in class results?"

[00:10–00:15]
Visual: Transition to a full-screen talking head of the creator in his room. Soft warm lighting, bookshelves in the background. Text overlay: "over the last 2 years".
Subject: Creator speaking directly to camera, gesturing with hands.
Speech: "Well I've been using them over the last 2 years and here is a..."

[00:15–00:20]
Visual: Fast montage of cinematic clips: A woman in a white dress in water with floating clothes ("3 best models"), a red-tinted close-up of a person in goggles ("that you can access"), a man in a hat walking in a foggy field ("under one subscription"). Text overlay: "FREEP!K".
Speech: "...no fluff, no BS list of the three best models that you can access under one subscription on Freepik."

[00:20–00:24]
Visual: Background shows a 1950s style dialogue scene between a man in a tweed suit and a woman in a beret (Veo 3.1).
Subject: Creator in circular overlay, thumbs up.
Speech: "Veo 3.1 is best for dialogue and lip-sync performance..."

[00:24–00:28]
Visual: Background shows a "Behind the scenes" shot of an Asian woman on a green screen set, then a "Fix" shot of a man being shaved with high skin detail. A red "X" and green "Checkmark" appear.
Subject: Creator explains the "plastic skin" issue.
Speech: "...but it can lead to plasticky skin textures. To avoid this, you can generate close-up shots and it'll give you better results."

[00:28–00:34]
Visual: Background shows a black and white shot of hands praying, then a fashion model against a white textured wall. The camera dollys in close to her eye, showing extreme detail. Text: "Kling 2.6".
Subject: Creator gesturing "dynamic" with hands.
Speech: "Kling 2.6 is the B-roll king. You can add in multiple camera directions into your prompt to get more dynamic results."

[00:34–00:38]
Visual: Background shows a man boxing a heavy bag, then a man lifting a heavy barbell in a gym. Text: "Hailuo 2.3".
Subject: Creator nodding.
Speech: "And Hailuo 2.3 is the best AI video model for complex movements."

[00:38–00:42]
Visual: Background shows the Freepik website UI scrolling through AI models. Large text overlay: "Comment AI".
Subject: Creator looking at the camera, smiling.
Speech: "You can test all of these on Freepik, so type AI in the comments and I'll send you a link."

NEGATIVE PROMPT: Visual artifacts, distorted limbs, flickering lighting, blurry faces in background, robotic lip-sync, inconsistent hat logo, low-resolution textures, harsh digital noise, unnatural eye movements, text clipping.

SPEECH PACK:
[00:00-00:10]
Transcript: "There's a lot of great AI video models out there. But which one should you be spending your hard-earned money on? Which one is the most cost-effective? And what's going to give you the best in class results?"
TAKE_A: (Energetic, fast-paced, questioning tone)
TAKE_B: (Authoritative, steady, emphasizing "hard-earned money")
TAKE_C: (Casual, conversational, friendly)

[00:10-00:20]
Transcript: "Well I've been using them over the last 2 years and here is a no fluff, no BS list of the three best models that you can access under one subscription on Freepik."
TAKE_A: (Confident, leaning in, emphasizing "no fluff")
TAKE_B: (Professional, clear enunciation of "Freepik")

[00:20-00:42]
Transcript: "Veo 3.1 is best for dialogue and lip-sync performance but it can lead to plasticky skin textures. To avoid this, you can generate close-up shots and it'll give you better results. Kling 2.6 is the B-roll king. You can add in multiple camera directions into your prompt to get more dynamic results. And Hailuo 2.3 is the best AI video model for complex movements. You can test all of these on Freepik, so type AI in the comments and I'll send you a link."
TAKE_A: (Instructional, helpful, clear transitions between model names)
TAKE_B: (Fast, punchy, direct-to-camera CTA)

Why rourke's Freepik Video Model Comparison Went Viral - and the Formula Behind It

This case study analyzes a high-performing "Tool Comparison" video by creator @rourke, which effectively breaks down the complex landscape of 2025 AI video models. The video uses a hybrid talking-head and cinematic B-roll format, featuring a central circular overlay of the creator superimposed on high-quality AI-generated footage. By comparing models like Veo 3.1, Kling 2.6, and Hailuo 2.3 through the lens of a single platform (Freepik), the creator solves a major pain point for indie creators: decision paralysis and subscription fatigue. The aesthetic is "Tech-Cinematic," blending clean UI elements (rating cards) with hyper-realistic AI visuals ranging from fashion editorials to gritty gym shots. This format is a masterclass in educational authority and engagement automation, leveraging the "Comment AI" hook to drive massive reach through Instagram's algorithm.

What You’re Seeing: Visual & Audio Analysis

The video is a fast-paced, information-dense edit. It begins with a "grid view" of various AI models, immediately establishing the scope. The creator, @rourke, appears in a circular mask with a soft glow, providing a consistent human anchor. The background shifts between specific AI-generated clips that demonstrate the strengths of each model: Veo 3.1 shows detailed dialogue and skin; Kling 2.6 showcases dynamic camera movements and macro eye shots; Hailuo 2.3 handles complex human motion like boxing and weightlifting.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:02	Grid of 8 AI video thumbnails with model names.	Wide UI layout, fast zoom.	High-contrast, vibrant.	Hook: Establish variety and authority.
00:02–00:10	Individual model "Rating Cards" (Cost, Speed, Quality).	UI Overlay over cinematic clips.	Clean, professional tech-vibe.	Value: Quick comparison metrics.
00:10–00:20	Creator talking head (MCU) in a circle.	Direct-to-camera, expressive.	Warm indoor practical lighting.	Trust: Humanizing the AI content.
00:20–00:28	Veo 3.1: Period drama dialogue + skin texture comparison.	Medium shots to extreme close-ups.	Cinematic, warm (drama) vs. Studio (BTS).	Education: Specific use-case (Dialogue).
00:28–00:34	Kling 2.6: Fashion model on snowy dunes + macro eye shot.	Dolly-in, extreme macro.	High-key, editorial, cool tones.	Aspiration: High-end B-roll quality.
00:34–00:38	Hailuo 2.3: Boxing and weightlifting in a gym.	Action tracking, handheld feel.	Moody, gym-lighting, high shadows.	Proof: Handling complex physics/motion.
00:38–00:42	Freepik UI scroll + "Comment AI" text overlay.	Screen recording + Text pop.	App-interface clean look.	Conversion: Clear Call-to-Action.

Why It Went Viral: The Breakdown

The "Curated Authority" Strategy

This video taps into the "Curator Economy." In a world where new AI tools are released weekly, users are overwhelmed. @rourke positions himself as the filter. By providing a "no-fluff, no-BS list," he provides immediate psychological relief to the viewer. The use of rating cards (Cost/Speed/Quality) mimics video game stats, which is a highly digestible and shareable format. It transforms a complex technical topic into a "cheat sheet" that users want to save for later.

Psychological Triggers & Platform Signals

The video utilizes Loss Aversion ("Which one should you be spending your hard-earned money on?") to grab attention in the first 3 seconds. From a platform perspective, the "Comment AI" call-to-action is the engine. Instagram rewards videos with high comment-to-view ratios. By promising a link via DM, the creator triggers an engagement loop: more comments → higher reach → more views → more comments. The fast pacing (cuts every 1.5–2 seconds) ensures high watch time, as the viewer has to stay focused to catch the specific pros and cons of each model.

5 Testable Viral Hypotheses

The "Comparison Card" Effect: Using standardized rating scales (dots/stars) for tools increases "Save" rates by 40% because it looks like a reference guide.
The Circular Human Anchor: Keeping a human face visible (even in a small overlay) during AI footage reduces the "uncanny valley" feeling and builds personal brand trust.
The "One-Stop-Shop" Narrative: Promoting a single platform that hosts multiple tools (Freepik) performs better than reviewing individual tools because it simplifies the user's workflow.
Visual Proof-of-Concept: Showing a "Fail" (plastic skin in Veo) followed by a "Fix" (close-up shots) creates a mini-tutorial arc that increases "Share" value.
The Keyword Trigger: Using a short, 2-letter comment trigger ("AI") lowers the friction for engagement compared to asking a complex question.

How to Recreate: From 0 to 1

Topic Selection: Choose a "Best of" or "Comparison" topic within your niche (e.g., "3 Best CapCut Features" or "Top 3 AI Image Generators").
Scripting for Retention: Start with a pain point (Money/Time). Use the "Pros/Cons/Fix" structure for each item you review.
Source AI Visuals: Generate 3 distinct styles of video. 1) Dialogue-heavy, 2) Cinematic B-roll, 3) High-motion action. Ensure they look premium.
Record Talking Head: Film yourself in a well-lit room. Wear a consistent "brand" outfit (like the Vans cap in the video). Keep your energy high.
Design UI Elements: Create "Rating Cards" in Canva or Figma. Use simple icons for Cost, Speed, and Quality.
The "Circle Overlay" Edit: In CapCut or Premiere, place your talking head on a top layer. Use a "Circle Mask" and add a slight "Outer Glow" or "Stroke" to make it pop from the background.
Syncing Cuts: Ensure the background video changes exactly when you mention a new model name. Use text overlays to reinforce the model names.
Automation Setup: Use a tool like ManyChat. Set a trigger: "If user comments 'AI', send them a DM with the link to the tools."
Cover Strategy: Use a high-contrast frame of the AI footage with a bold title: "BEST AI VIDEO MODELS 2025."

Growth Playbook: Distribution & Scaling

3 Opening Hook Lines

"Stop paying for 5 different AI video subscriptions. Do this instead."
"I tested every AI video model in 2025 so you don't have to."
"The secret to cinematic AI video isn't the prompt, it's the model."

4 Caption Templates

The "Value Bomb" Template:
Which AI video model is actually worth your money? 💸

I’ve spent 2 years testing these tools, and the landscape just changed. Here’s the breakdown:
- Veo 3.1: Best for [Value]
- Kling 2.6: Best for [Value]
- Hailuo 2.3: Best for [Value]

Which one are you trying first? 👇
Comment "AI" and I'll send you the access link!

Hashtag Strategy

Broad (Reach): #aivideo #generativeai #creativeai #techtrends
Mid-Tier (Niche): #freepik #videoproduction #contentcreator #aiart
Long-Tail (Specific): #klingai #veo3 #hailuoai #aivideotutorial

FAQ

What tools make the circular overlay look the most professional?

Use CapCut's "Mask" tool set to "Circle" and add a white "Stroke" or "Glow" effect.

How do I keep the AI video quality high after uploading to Instagram?

Export in 1080p, 30fps, with a high bitrate (at least 15-20 Mbps) and enable "High Quality Uploads" in IG settings.

Why does the creator use a circular mask instead of a full screen?

It allows the viewer to see the "Proof" (the AI video) while still connecting with the "Teacher" (the creator).

Is it better to use ManyChat or reply manually?

For videos with over 100 comments, ManyChat is essential to avoid being flagged for spam by Instagram.

How can I avoid making AI video look "plastic"?

As mentioned in the video, use close-up prompts to force the AI to render more skin texture and pores.

What is the most important word in the prompt for these models?

"Cinematic," "Texture," and specific camera movements like "Dolly-in" are crucial for these results.