0:00 / 0:00

Ylenia Fedi

@yleintech

INSTAGRAM · 2026-03-19Source

144likes

446comments

Remix This

Recreate with Runway Gen 4.5

Make your own AI viral video

Prompt

GLOBAL LOCK: 
Subject: A young woman in her mid-20s, light skin with warm undertones, long wavy dark brown hair parted in the middle. She wears a white ribbed turtleneck sweater and a silver watch on her left wrist. 
Environment: A clean studio with a soft purple and pink gradient background. A dark desk and the edge of a laptop are visible in the foreground.
Style: High-definition UGC tech tutorial, clean lighting, vibrant colors.
AI Animation Style: High-fidelity 3D cartoon animation (saturated colors, smooth motion) and cinematic photorealistic action.
Speech: Female voice, enthusiastic and clear, medium pace, professional mic quality with slight room resonance.

[00:00–00:02]
Subject: Host looking directly at camera, speaking.
B-roll Overlay: A cinematic, high-speed desert buggy racing through sand dunes, massive dust clouds billowing behind it. High-contrast, bright sunlight.
Action: Host gestures slightly with hands. Buggy moves rapidly from left to right.
Speech: "There's a website where you can create"
Sync: High lip-sync strictness.

[00:02–00:04]
Subject: Host speaking.
B-roll Overlay: Close-up of the buggy's wheels churning sand, intense motion blur. Text "Consistent AI Videos" appears in bold yellow.
Action: Fast-paced action shot.
Speech: "consistent AI videos like Higgsfield AI"
Sync: Cut lands on "Higgsfield".

[00:04–00:06]
Subject: Host speaking, smiling.
B-roll Overlay: Higgsfield AI logo (green square with a black squiggle).
Speech: "and it's completely free."
Sync: High lip-sync strictness.

[00:06–00:09]
Subject: Host speaking.
B-roll Overlay: Screen recording of the Higgsfield interface. A panda is shown in a video preview. Text "Just paste or prompt" appears.
Action: Mouse cursor hovers over the "Create Video" button.
Speech: "Just paste your image or prompt and the platform"

[00:09–00:13]
Subject: Host speaking.
B-roll Overlay: A scrolling list of AI models (Claude, Gemini, Grok) followed by a grid of AI video tool logos.
Action: Rapid scrolling motion.
Speech: "generates the video for you. Now you might think every AI video tool can do that,"

[00:13–00:17]
Subject: Host speaking, leaning forward slightly.
B-roll Overlay: Screen recording of a 3D cartoon cat chasing a mouse. A right-click menu appears over the video.
Action: Mouse selects "Copy Video Frame".
Speech: "but here's what makes this one special. After the video is generated,"

[00:17–00:20]
Subject: Host speaking.
B-roll Overlay: The copied frame is pasted into the prompt box.
Action: UI interaction showing the image being uploaded.
Speech: "you can right-click and copy the last frame, then paste that frame"

[00:20–00:26]
Subject: Host speaking (small window) / Full-screen animation.
Visual: The 3D cartoon cat continues the chase, running through a hole in the wall. The mouse is seen inside the wall with a piece of cheese.
Action: Smooth, high-speed character animation. The cat looks frustrated.
Speech: "back into the tool and continue the prompt. The AI will continue the story exactly where the first video ended,"

[00:26–00:31]
Subject: Host speaking.
B-roll Overlay: A new animation of a stylized 3D family (father, mother, two children) standing outside a house. A yellow school bus drives into the frame.
Action: The bus stops, the camera pans slightly.
Speech: "keeping the same characters and visual style. So instead of random clips, you can create consistent story-based videos"

[00:31–00:34]
Subject: Host speaking directly to camera, friendly expression.
Visual: Text "Comment Video" and "send you Video" appears in yellow.
Action: Host clasps hands on the desk.
Speech: "scene after scene. Want to try it yourself? Comment 'Video' and I'll send you over."
Sync: High lip-sync strictness on CTA.

NEGATIVE PROMPT: Visual artifacts, distorted faces, flickering backgrounds, inconsistent clothing colors, robotic mouth movements, blurry UI text, harsh shadows on the host, unnatural hair physics in animation, audio clipping, background noise, muffled speech.

SPEECH PACK:
[00:00-00:04] "There's a website where you can create consistent AI videos like Higgsfield AI"
TAKE_A: (Enthusiastic, fast) "There's a website where you can create consistent AI videos like Higgsfield AI"
TAKE_B: (Informative, steady) "There's a website... where you can create consistent AI videos... like Higgsfield AI"

[00:04-00:13] "and it's completely free. Just paste your image or prompt and the platform generates the video for you. Now you might think every AI video tool can do that,"
TAKE_A: (Emphasizing 'free' and 'every') "and it's completely FREE. Just paste your image or prompt and the platform generates the video for you. Now you might think EVERY AI video tool can do that,"

[00:13-00:20] "but here's what makes this one special. After the video is generated, you can right-click and copy the last frame, then paste that frame"
TAKE_A: (Intriguing tone) "but here's what makes THIS one special. After the video is generated, you can right-click and copy the last frame, then paste that frame"

[00:20-00:34] "back into the tool and continue the prompt. The AI will continue the story exactly where the first video ended, keeping the same characters and visual style. So instead of random clips, you can create consistent story-based videos scene after scene. Want to try it yourself? Comment 'Video' and I'll send you over."
TAKE_A: (Helpful and encouraging) "back into the tool and continue the prompt. The AI will continue the story EXACTLY where the first video ended... keeping the same characters and visual style. So instead of random clips, you can create consistent story-based videos scene after scene. Want to try it yourself? Comment 'Video' and I'll send you over!"

How yleintech Made This Higgsfield AI Consistent Video Tutorial — and How to Recreate It

This case study analyzes a high-performing tech tutorial by @yleintech, focusing on Higgsfield AI, a tool designed to solve the "consistency" problem in AI video generation. The video features a clean, UGC-style studio aesthetic with a soft purple gradient background, high-quality 3D animations (reminiscent of Pixar and classic cartoons), and a clear, actionable UI walkthrough. By demonstrating a specific "copy-last-frame" workflow, the creator provides immediate value to indie filmmakers and content creators struggling with character continuity in AI tools like Runway or Luma.

What You’re Seeing

The video alternates between a "talking head" host and high-energy B-roll. The host is a young woman with long, wavy brown hair, wearing a professional yet accessible white turtleneck. The lighting is soft and directional, creating a flattering "beauty" look. The B-roll transitions from a cinematic, high-octane desert buggy race (demonstrating motion capabilities) to a charming 3D cartoon chase between a cat and a mouse (demonstrating character consistency).

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:03	Host intro + Desert buggy racing B-roll	Medium shot / Fast-cut action	Warm, high-contrast desert sun	Hook: High-quality visual proof
00:04–00:09	Host + Higgsfield website UI	Medium shot / Screen recording	Soft studio purple / Clean UI	Introduce the "Free" solution
00:10–00:13	List of AI models + Tool logos	Fast-paced graphic overlay	Dark mode UI / Vibrant logos	Establish authority & context
00:14–00:19	UI Demo: Right-clicking last frame	Close-up screen recording	High clarity / Instructional	The "Aha!" moment (The Secret)
00:20–00:25	Cat chasing mouse (Part 2)	3D Animation / Side-scrolling	Vibrant, saturated cartoon colors	Proof of consistency mechanism
00:26–00:30	Family & School Bus animation	Cinematic 3D render	Golden hour / Warm & Emotional	Expand use-case to storytelling
00:31–00:34	Host CTA (Comment "Video")	Medium shot / Direct eye contact	Soft studio purple	Conversion: Drive engagement

Why It Went Viral

The Problem-Solver Strategy

The core of this video's success is its focus on a universal pain point: character consistency in AI video. Most creators are frustrated by AI's tendency to change a character's face or clothes between shots. By positioning Higgsfield AI as the "fix" for this, the creator taps into a high-intent audience. The "completely free" mention acts as a low-friction entry point, making the value proposition irresistible.

Platform Signals & Psychological Triggers

From a platform perspective, the video uses a "Comment for Link" strategy. This is a powerful engagement hack. Instead of putting a link in the bio (which platforms often deprioritize), the creator forces users to comment. This signals to the algorithm that the content is highly engaging, triggering a massive boost in reach. The fast-paced editing and high-quality B-roll ensure high retention (watch time), while the tutorial nature encourages "saves" for future reference.

5 Testable Viral Hypotheses

The "Secret Feature" Hook: Showing a specific, non-obvious UI action (right-click -> copy frame) creates a sense of "insider knowledge" that viewers feel compelled to save.
The Contrast Effect: Switching from a high-octane desert race to a cute cartoon demonstrates the tool's versatility, appealing to both "tech bros" and "creative storytellers."
The "Free" Multiplier: In a market saturated with expensive subscriptions, highlighting a "completely free" tool significantly lowers the barrier to sharing.
Visual Proof of Concept: Using a "Tom & Jerry" style chase is a brilliant choice; everyone understands the physics and character requirements of a chase scene, making the "consistency" claim easy to verify visually.
The Engagement Loop: The CTA "Comment Video" creates a feedback loop where every comment pushes the video to a new set of users, creating a viral snowball effect.

How to Recreate

Identify a "Consistency" Tool: Choose an AI tool (like Higgsfield, Kling, or Luma) that has a specific feature for character or frame continuity.
Script the "Aha!" Moment: Your script should lead with the problem (inconsistent AI) and quickly reveal the specific button or workflow that solves it.
Record High-Quality B-Roll: Generate 2-3 distinct styles of video (e.g., cinematic, 3D animation, claymation) to show the tool's range.
Film Your Talking Head: Use a clean background. A soft, single-color or gradient background works best to keep the focus on you and the overlays.
Screen Record the UI: Don't just talk about the tool; show the exact mouse movements. Use a zoom-in effect on the specific buttons you mention.
Edit for Retention: Use text overlays for every key phrase. Ensure there is a visual change (cut or overlay) every 2-3 seconds.
Set Up Automation: Use a tool like ManyChat to automatically send the link to anyone who comments your keyword.
Optimize for the Platform: Use trending but relevant audio at a low volume (5-10%) to help the algorithm categorize your content.

Growth Playbook

Opening Hook Lines

"Stop struggling with inconsistent AI characters. Use this instead."
"The secret to making full AI movies is finally free."
"I found the only AI video tool that actually keeps your characters the same."

Caption Templates

The "Value First" Template:
Character consistency is the #1 killer of AI films. 🎬 But I found a workaround. [Tool Name] lets you copy the last frame to continue your story perfectly.
Value: No more changing faces or outfits.
Question: What kind of AI story are you dreaming of making?
CTA: Comment "LINK" and I'll send you the tool!

Hashtag Strategy

Broad: #AI #ArtificialIntelligence #TechTrends #ContentCreator
Mid-Tier: #AIVideo #VideoEditing #Filmmaking #AITools
Niche: #HiggsfieldAI #CharacterConsistency #AIFilm #IndieCreator

Frequently Asked Questions

What tools make it look the most similar?

Higgsfield AI is the primary tool, but Luma Dream Machine and Kling AI also offer strong frame-to-video features.

What are the 3 most important words in the prompt?

"Consistent," "Character," and "Continuation" are key for this specific workflow.

Why does the generated face look inconsistent?

Usually, it's because the "seed" or the reference frame isn't strong enough; always use the "copy last frame" method shown.

How can I avoid making it look like AI?

Use high-quality 3D animation styles (like Pixar) which are more forgiving than photorealistic human faces.

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

Instagram Reels currently has a higher "save" rate for educational tech content like this.