0:00 / 0:00

Ojito con Soul Cinema 🤩🎥 Comenta “SOUL” y te envío los prompts 📨🔥 El modelo de imágenes de @higgsfield.ai recibe una actualización y ahora desde Soul Cinema se pueden crear escenas realmente cinematográficas. En estos videos te muestro 10 ejemplos que parten de imágenes creadas con Soul Cinema y y Nano Banana 2. Mismos prompts con ambos modelos. Sinceramente me sorprendió bastante lo bien que lo hace este modelo para crear este tipo de escenas. Como punto débil hay que decir que no tiene una gran adherencia al prompt en comparación con NB2. Además, con Soul ID y Soul HEX se pueden crear imágenes con nuestro rostro de forma súper precisa y controlar el color de las imágenes entre las escenas, respectivamente. Muy interesante este modelo de Higgsfield 👌🏽 Por cierto, dos cosas más. Esta comparación deja mal a NB2, pero esto seguro que con el prompt adecuado se pueden conseguir imágenes muy top. Y en segundo lugar, todas las imágenes fueron animadas con Kling 3.0… qué modelo de vídeo más bestia 🥹

Pablo Prompt

@pabloprompt

INSTAGRAM · 2026-03-07Source

570likes

125comments

Remix This

Recreate with Kling 3

Make your own AI viral video

Prompt

GLOBAL LOCK: A group of four diverse friends (two Black women, one Caucasian woman, one Black man) in their early 20s. They are dressed in casual 90s-inspired denim outfits: denim jackets, light-wash jeans, and white t-shirts. The setting is a rooftop parking lot at sunset with a hazy city skyline in the distance. The lighting is warm "Golden Hour" with strong backlighting, creating rim light on hair and soft lens flares. The color grade is cinematic with warm oranges and deep blues. The camera has a slight handheld jitter for a realistic feel.

[00:00–00:03]
The group is leaning against the back of a dark grey hatchback car with the trunk open. The woman on the far left is throwing her head back in a deep, genuine laugh. The woman in the center is clapping her hands together joyfully. The man on the right is smiling and looking at his friends. Wide shot showing the car and the city horizon. High-fidelity motion, hair blowing slightly in the breeze.

[00:03–00:06]
Medium close-up on the three women. The central woman with curly hair is leaning forward, laughing intensely, her shoulders shaking. The woman to her left has her eyes closed in laughter. The lighting is very warm, catching the edges of their denim jackets. The camera pans slightly to the right.

[00:06–00:10]
The man on the right reaches out a hand to pat the shoulder of the woman next to him. The group continues to laugh and interact. The sun is lower now, creating a more dramatic orange glow across the scene. The city lights in the background begin to twinkle. The motion is fluid and natural, capturing the micro-expressions of joy.

NEGATIVE PROMPT: Robotic movement, frozen faces, distorted limbs, flickering lighting, blurry textures, inconsistent clothing, morphing backgrounds, low resolution, watermarks, text, cartoonish style, unnatural skin tones.

SPEECH PACK:
[00:00-00:10]
Transcript: "[Laughter] ... That is so funny! ... [Laughter]"
TAKE_A: High-pitched, energetic group laughter with a clear "That is so funny!" in the middle.
TAKE_B: More wheezing, breathless laughter with a muffled "Oh my god" instead of the main line.
TAKE_C: Relaxed, chuckling laughter with a very clear, enunciated "That is so funny!" at the 7-second mark.
Prosody: Natural pauses for breath, overlapping voices, warm and friendly tone.
Sync: High lip-sync strictness for the "That is so funny" line if the central woman is on camera.

How pabloprompt Made This AI Video Comparison Higgsfield Soul Cinema — and How to Recreate It

This viral case study examines a high-engagement AI model comparison video by @pabloprompt. The video features a split-screen "battle" between two AI video generation engines: Higgsfield Soul Cinema and Nano Banana 2. The scene depicts a group of four friends sharing a genuine moment of laughter on a rooftop at sunset, leaning against a car with a city skyline in the background. With its warm, cinematic "Golden Hour" aesthetic and high-fidelity motion, the video serves as both a technical showcase and a piece of aspirational lifestyle content. It leverages the "Comparison" format—a proven viral hook in the AI community—to drive comments and saves from creators looking for the next best tool.

What You’re Seeing

The video is a vertical split-screen. The top half showcases "Soul Cinema," focusing on tight, intimate medium shots with rich, warm color grading and significant lens flare. The bottom half, "Nano Banana 2," provides a wider perspective, showing the full car and the expansive urban horizon. The subjects are a diverse group of young adults in casual denim attire, exhibiting fluid, naturalistic movements like clapping, head-tilting, and expressive laughter.

Shot-by-Shot Breakdown (Estimated)

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00 - 00:03	Friends laughing by the car.	Medium Shot (Top) / Wide Shot (Bottom)	Golden hour, high contrast, warm oranges.	Hook: Immediate visual "wow" factor.
00:03 - 00:06	Subject in center claps and leans back.	Dynamic motion, handheld feel.	Rim lighting on hair, soft shadows.	Demonstrate motion consistency and realism.
00:06 - 00:10	Group interaction, city lights appearing.	Static wide (Bottom) vs. tight (Top).	Sunset transition, cinematic teal/orange.	Reinforce tool capability and "vibe."

Why It Went Viral

The Power of Comparison

This video taps into the "Tool Comparison" psychology. In a rapidly evolving field like AI, creators are constantly searching for the "best" model. By placing two outputs side-by-side, the creator forces the viewer to judge and choose a favorite. This naturally leads to high engagement in the comments as users debate which model looks more "real" or "cinematic."

Aesthetic Aspiration (The Golden Hour Effect)

The choice of a sunset rooftop setting isn't accidental. It utilizes biological triggers—warm light and human laughter—to create an immediate positive emotional response. The "Soul Cinema" model specifically emphasizes skin textures and light wrap, which are high-value signals for "cinematic" quality that indie filmmakers crave.

Platform Perspective & Signals

From an Instagram/TikTok perspective, this video excels due to:

Loopability: The laughter and ambient noise create a seamless loop, boosting watch time.
Saveability: Creators save this as a reference for their own AI prompts or to remember the tool name (Higgsfield).
Comment Bait: The caption "Comenta SOUL" (Comment SOUL) uses a keyword-triggered automation to send prompts, skyrocketing the comment count and signaling to the algorithm that the content is highly interactive.

5 Testable Viral Hypotheses

The Side-by-Side Test: Comparing a known tool with a new "challenger" increases curiosity.
The "Humanity" Factor: AI videos showing genuine laughter perform better than static or "cool" poses because they are harder to generate convincingly.
The Keyword Trigger: Using "Comment [Word]" for a DM reward significantly boosts reach compared to a standard CTA.
Golden Hour Dominance: Warm, backlit scenes consistently receive higher "aesthetic" ratings from viewers.
Text Overlay Clarity: Using bold, clear labels for the tools prevents confusion and keeps the viewer focused on the comparison.

How to Recreate: Step-by-Step AI Workflow

Topic Selection: Choose two AI video models (e.g., Luma Dream Machine vs. Kling, or Higgsfield vs. Runway).
Scene Concept: Design a scene with high "motion difficulty"—laughter, hair blowing in the wind, or complex lighting.
Prompt Engineering: Use a consistent base prompt for both models. Focus on keywords like "cinematic sunset," "rooftop parking lot," and "candid laughter."
Character Consistency: If the model supports it, use a "Character Reference" (Cref) image to ensure the friends look the same across different generations.
Generation: Generate multiple 5-10 second clips. Look for the ones with the least "AI warping" in the faces during laughter.
Editing: Use CapCut or Premiere Pro to create a vertical split-screen. Add text overlays at the top and bottom to label the models.
Audio Layering: Add high-quality ambient rooftop sounds and a subtle "laughing group" audio track to mask any AI-generated audio artifacts.
Engagement Hook: Write a caption offering the "Secret Prompts" in exchange for a specific comment word.

Growth Playbook

Opening Hook Lines

"Is [Model A] finally better than [Model B]? Watch this."
"I tested the new Higgsfield update so you don't have to."
"Stop using boring AI video. This new model is a game changer."

Caption Templates

The "Tool Reveal" Template:
The AI video wars are heating up! 🎥🔥
I compared Soul Cinema vs. Nano Banana 2 and the results are insane. One is clearly more cinematic, but the other handles wide shots better.
Which one do you prefer? 👇
Comment "PROMPT" and I'll send you the exact settings I used! 📨

Hashtag Strategy

Broad: #aivideo #artificialintelligence #filmmaking #creativetech
Mid-tier: #higgsfield #runwayml #lumadreammachine #digitalcreator
Niche: #aiprompttips #cinematicai #indiecreators #pabloprompt

Frequently Asked Questions

What tools make it look the most similar?

Higgsfield and Luma Dream Machine are currently best for this specific warm, cinematic look.

What are the 3 most important words in the prompt?

"Golden hour," "candid laughter," and "backlit."

Why does the generated face look inconsistent?

Extreme expressions like wide laughter often break AI facial structures; try lowering the "motion" slider.

How can I avoid making it look like AI?

Add a slight film grain and handheld camera shake in post-production to mimic real cinematography.

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

Instagram Reels currently favors high-aesthetic "vibe" content like this comparison.