How to keep characters consistent?

Use the --cref parameter in Midjourney.

▶

One night in New York❄️ Are you serious about learning how to create AI visuals at a professional level? In my course, I teach everything I use in my own process: • How I build concepts and visual storytelling • The exact tools, prompts, and workflows I use • How to refine, edit, and create a portfolio that gets noticed This is the full system I used to grow Dreamfall Art and work with brands. Comment COURSE for the link. Animated with @higgsfield.ai #aiart #ai #midjourney #newyork #retro #aiartwork #newyorkcity #manhattan

dreamfall.art

@dreamfall.art · creator

INSTAGRAM · 2025-11-25Source

art AI newyork retro AIart midjourney

83.9Klikes

500comments

Remix This

Prompt

GLOBAL LOCK: A sophisticated Caucasian couple in their late 20s. The woman has dark brown hair in an elegant updo, wearing a black backless evening gown with a high slit and silver earrings. The man is clean-cut with short dark hair, wearing a classic black tuxedo with a white shirt and black bowtie. The setting is a luxury NYC interior at night, featuring floor-to-ceiling windows with a view of the Empire State Building. Heavy snow is falling outside. Lighting is warm candlelight inside and cool blue-toned city lights outside. Cinematic 35mm film style, soft grain, high contrast, elegant and romantic mood. No speech, ambient romantic orchestral music.

[00:00–00:03]
Medium shot of the couple sitting at a dinner table. The woman is facing the camera, smiling softly, while the man is seen in partial profile looking at her. The Empire State Building is perfectly framed in the window behind them. Snowflakes are falling steadily outside. Warm candle glow on their faces.

[00:03–00:06]
Medium shot from behind the woman as the couple slow dances. The focus is on the intricate backless detail of her black dress and her elegant posture. The man's hands are on her waist. The background is a blur of golden fairy lights and the dark NYC skyline.

[00:06–00:08]
Over-the-shoulder shot from behind the couple. The woman points her finger toward the Empire State Building through the glass. The camera captures the reflection of the interior lights on the window pane. The snow creates a dreamy, hazy atmosphere outside.

[00:08–00:11]
Close-up shot of the couple. They are holding champagne glasses, bringing them together for a toast. They lean in close, their foreheads touching, and share a gentle, romantic kiss. Shallow depth of field, blurring the city lights into golden bokeh.

[00:11–00:15]
Wide shot of the couple dancing in front of the massive window. The room is decorated with warm fairy lights along the window sill. The entire Manhattan skyline is visible, covered in a winter snowstorm. The camera is static, capturing the timeless, cinematic beauty of the moment.

NEGATIVE PROMPT: low resolution, blurry faces, distorted anatomy, extra limbs, cartoonish, bright neon colors, daytime, no snow, casual clothing, messy hair, shaky camera, digital artifacts, flickering lights, text, logos, watermark, inconsistent character features.

SPEECH PACK:
(Note: This video contains no speech, only music.)
TRANSCRIPT: [Silent / Ambient Music]
DELIVERY: N/A
PROSODY: N/A
SYNC: N/A

Why dreamfall.art's NYC Romance AI Video Went Viral

This case study analyzes a viral cinematic AI video by @dreamfall.art, capturing a romantic "One night in New York" aesthetic. The video leverages the high-performing "Old Money" / "Quiet Luxury" niche, combining hyper-realistic AI characters with an iconic, aspirational setting: a candlelit dinner overlooking a snowy Manhattan skyline. With over 83k likes, it demonstrates how creators can use AI to bypass the massive production costs of high-end fashion shoots while maintaining a premium, editorial feel that resonates deeply on platforms like Instagram and TikTok.

What You’re Seeing: A Visual Breakdown

The video is a montage of romantic moments between a sophisticated couple. The subject is a woman in a plunging, backless black evening gown with a high slit, accompanied by a man in a classic black tuxedo. The scene is set in a luxury high-rise interior, likely a restaurant or penthouse, with floor-to-ceiling windows framing the Empire State Building.

The lighting is masterfully handled, featuring a "warm vs. cool" contrast: the interior is bathed in soft, golden candlelight and fairy lights (warm), while the exterior city lights and falling snow provide a crisp, blue-toned backdrop (cool). The texture mimics 35mm film with subtle grain and a soft highlight roll-off, making the AI-generated content feel grounded and "expensive." The music is a slow, romantic orchestral piece that dictates the slow-motion editing rhythm.

Shot-by-Shot Analysis

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:03	Couple at a dinner table; woman looks at the camera.	Medium Shot (MS), Eye-level.	Warm candlelight, bokeh city lights.	The Hook: Establishes the "Main Character" fantasy.
00:03–00:06	Couple dancing; focus on the woman's backless dress.	Medium Close-Up (MCU), tracking movement.	Golden rim lighting on the dress.	Aesthetic Value: Showcases fashion and elegance.
00:06–00:08	Woman points at the Empire State Building.	Over-the-shoulder (OTS) looking out.	High contrast; silhouette against the city.	Location Flex: Connects the viewer to the NYC dream.
00:08–00:11	Close-up of a toast and a gentle kiss.	Close-Up (CU), shallow depth of field.	Soft, romantic glow.	Emotional Connection: Peaks the romantic narrative.
00:11–00:15	Wide shot of dancing by the window with fairy lights.	Wide Shot (WS), static.	Warm fairy lights vs. cold snow.	The Loop: A final atmospheric shot that invites rewatching.

Why It Went Viral: The "Old Money" Psychology

The success of this video lies in its perfect alignment with the "Aspirational Lifestyle" trigger. By choosing New York City during a snowfall—a globally recognized symbol of romance and success—the creator taps into a collective cultural fantasy. The "Old Money" aesthetic (tuxedos, evening gowns, fine dining) triggers a psychological desire for status and elegance, which consistently performs well in the Instagram "Explore" feed.

From a platform perspective, the video excels in "Watch Time" and "Saves." The 0–3 second hook uses a high-quality, direct-to-camera gaze from a beautiful subject, which stops the scroll. The pacing is slow and rhythmic, matching the music, which creates a "trance-like" viewing experience. Users save these videos as "mood board" inspiration for photography, fashion, or travel goals, signaling to the algorithm that the content is high-value.

5 Testable Viral Hypotheses

The "Iconic Landmark" Anchor: Using a recognizable landmark (Empire State Building) increases shareability because it provides a clear "vibe" that people want to associate with their own profiles.
The "Warm/Cold" Color Contrast: The visual tension between the cozy interior and the freezing exterior creates a sensory "feeling" that keeps viewers engaged longer than monochromatic scenes.
The "Backless Dress" Reveal: High-end fashion details (like the backless gown at 0:04) act as a secondary hook for fashion-conscious audiences, increasing "Saves."
The "Snowfall" Loop: Continuous, gentle motion (like falling snow) makes the video feel alive and seamless, encouraging multiple loops without the viewer realizing it.
The "Silent Romance" Narrative: By not having dialogue, the video becomes universal. Anyone, regardless of language, can project themselves into the romantic scene.

How to Recreate: From 0 to 1

1. Topic Selection & Positioning

This style suits accounts focused on AI Art, Luxury Lifestyle, Fashion Inspiration, or Travel Romance. Your goal is to sell a "feeling" rather than a product.

2. Character Consistency

To keep the couple looking the same across shots, use a "Character Reference" (Cref) in Midjourney. Generate one high-quality portrait of the woman and one of the man first. Use these as your base for all subsequent scene generations.

3. Prompting the Vibe

Focus on lighting and material keywords: "cinematic lighting," "candlelit," "35mm film grain," "silk evening gown," "snowing outside window."

4. Video Generation (The Motion)

Use tools like Luma Dream Machine, Runway Gen-3, or Kling AI. Upload your Midjourney images and use "End Frames" or "Motion Brushes" to ensure the snow falls and the couple moves naturally (e.g., slow dancing, toasting).

5. Adding the Snow Effect

If the AI doesn't generate enough snow, use a "Snow Overlay" in CapCut or Premiere Pro. Set the blending mode to "Screen" to make the snowflakes pop against the dark city background.

6. Color Grading for "Old Money"

Apply a "Teal and Orange" grade but desaturate the oranges slightly to keep it looking sophisticated. Increase the "Glow" or "Bloom" on the highlights (the candles and city lights).

7. Sound Design

Choose a trending "Cinematic" or "Classical" audio track. Ensure the cuts in your video land on the "downbeat" of the music for a professional feel.

8. Publishing Strategy

Post as a Reel/TikTok with a short, evocative caption. Do not over-explain. Let the visuals do the heavy lifting.

Growth Playbook: Distribution & Scaling

Opening Hook Lines

"POV: You're spending a snowy night in NYC. ❄️"
"The AI aesthetic that's taking over Instagram."
"Manifesting this 'Old Money' winter romance."

Caption Templates

Option 1 (Emotional):
One night in New York. ❄️ Sometimes the most beautiful moments are the ones we dream up. Which city should I do next? #AIArt #NYCWinter #OldMoney

Option 2 (Educational):
How I created this cinematic NYC scene using AI. 🎥 It’s all about the lighting contrast. Full prompt breakdown in my bio! #AIContentCreator #RunwayGen3 #CinematicAI

Hashtag Strategy

Broad: #AIArt #Cinematic #Travel #Romance #WinterVibes
Mid-tier: #OldMoneyAesthetic #NYCStyle #DigitalArt #LuxuryLifestyle
Niche: #MidjourneyV6 #RunwayGen3 #AIModel #DreamyAesthetic

Frequently Asked Questions

What tools make it look the most similar?

Midjourney for the base images and Runway Gen-3 or Kling AI for the high-fidelity video motion.

What are the 3 most important words in the prompt?

"Cinematic," "Candlelit," and "Photorealistic."

Why does the generated face look inconsistent?

You need to use a consistent "Character Reference" image and keep your prompt descriptions of the face very simple.

How can I avoid making it look like AI?

Add film grain in post-production and avoid "over-smoothing" the skin in your prompts.

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

Instagram, as the "Aesthetic/Lifestyle" niche is currently more dominant there for high-end visuals.