Which AI tool is best?

Runway Gen-3 or Luma Dream Machine are recommended for high-fidelity character work.

i figured out how to get perfect ai video generations first try while cutting costs by 60% for 6 months i was basically gambling and burning credits on attempts that were pure luck same setup, completely random results... sometimes perfect, usually trash, zero way to know which https://t.co/e8001EOS0f

Miko

@Mho_23 · creator

TWITTER · 2025-10-13Source

725likes

775comments

Remix This

Prompt

GLOBAL LOCK:
Subject: An elegant elderly woman (70s), Caucasian, with voluminous wavy white hair styled in a sophisticated bob. She wears thick black cat-eye sunglasses in every shot. Skin is fair with natural age-related texture but well-maintained. 
Wardrobe: Shot 1-2: White silk pajamas with black piping. Shot 3-6: High-end black lace dress with intricate floral patterns, silver hoop earrings, and a large emerald ring.
Environment: Luxury New York settings (Gold-toned bedroom, marble bathroom, black SUV interior, NYFW runway).
Lighting: High-end editorial lighting, warm and expensive.
Color Grade: Rich, high contrast, warm highlights, deep blacks, cinematic film grain.
Camera: Mix of handheld "vlog" style and professional "over-the-shoulder" interview framing.
Speech: Raspy, sophisticated, sassy female voice. Lip-sync must be high-precision for direct-to-camera lines.

[00:00–00:06]
Subject: Close-up of the woman in a gold-toned luxury bed. She holds a martini glass with an olive. A blurred 20-year-old man sleeps in the background.
Action: She speaks directly to the camera with a sassy smirk, takes a sip of the martini.
Camera: Handheld, slight "vlog" shake, 35mm lens.
Lighting: Soft morning light through a window, warm gold tones.
Speech: "What's up hoes? It's New York Fashion Week so obviously I woke up hungover as shit, but don't worry, I have just the cure."

[00:06–00:12]
Subject: Full shot of the woman in a massive, opulent marble bathroom. She holds the martini glass.
Action: She walks toward the camera, gesturing with her free hand, looking confident.
Camera: Wide angle, static but with slight organic movement.
Lighting: Bright, high-key bathroom lighting, multiple practical light sources (sconces).
Speech: "You guys keep saying I'm living too hard for my age. Why don't you ask the 20-year-old snoring in my ten thousand dollar sheets?"

[00:13–00:19]
Subject: Medium close-up inside a luxury black SUV.
Action: She looks out the window, then turns to the camera with a knowing smile, holding the martini.
Camera: MCU, side-profile to front-facing.
Lighting: Natural daylight from the car window, high contrast shadows.
Speech: "Money can't buy happiness, huh? Tell that to my front row fucking seat at the Khaite show."

[00:20–00:27]
Subject: Medium shot on a red carpet. She is being interviewed by a man in a suit holding a microphone with "NYFW" branding.
Action: She gestures to her dress and face while laughing. Paparazzi flashes in the background.
Camera: Over-the-shoulder, professional broadcast style.
Lighting: Cool, bright flash photography lighting.
Speech: "My dress is custom Scarlett, Khaite heels, and a facelift tighter than my prenup. Just kidding!"

[00:28–00:34]
Subject: Fashion runway. Models in oversized grey and black suits walk past a seated audience.
Action: Models walking with a neutral expression. The audience is clapping.
Camera: Low angle, runway-end perspective.
Lighting: Dramatic overhead spotlights, dark audience area.
Grade: Desaturated, high-fashion editorial look.

[00:35–00:45]
Subject: Close-up backstage. Clothes racks and models in the background.
Action: The woman holds the camera "selfie-style," looking excited. She pans to a silver-haired man in a black outfit walking the runway.
Camera: Handheld, shaky, wide-angle selfie lens.
Lighting: Harsh industrial backstage lighting.
Speech: "You guys won't believe it, I pulled a few strings and got my bestie Elliot to close the show!"

NEGATIVE PROMPT:
Visual: Cartoonish features, inconsistent hair color, sunglasses disappearing, martini glass merging with hand, blurry face, distorted background models, flickering lights, low resolution, plastic skin texture.
Speech: Robotic tone, monotone delivery, lip-sync lag, muffled audio, background noise overpowering the voice, unnatural pauses.

SPEECH PACK:
[00:00-00:06]
TAKE_A: "What's up hoes? It's New York Fashion Week so obviously I woke up hungover as shit... but don't worry, I have just the cure." (Sassy, raspy, playful)
[00:06-00:12]
TAKE_A: "You guys keep saying I'm living too hard for my age. Why don't you ask the 20-year-old snoring in my ten thousand dollar sheets?" (Defiant, wealthy tone)
[00:13-00:19]
TAKE_A: "Money can't buy happiness, huh? Tell that to my front row fucking seat at the Khaite show." (Sarcastic, sharp)
[00:20-00:27]
TAKE_A: "My dress is custom Scarlett, Khaite heels, and a facelift tighter than my prenup. Just kidding!" (Laughing, charismatic)
[00:35-00:45]
TAKE_A: "You guys won't believe it, I pulled a few strings and got my bestie Elliot to close the show!" (Excited, high energy)

How Mho_23 Made This AI Fashion Grandma AI Video

This viral case features "Darlene," a fictional AI-generated fashion icon who blends the "sassy grandma" trope with high-end New York Fashion Week (NYFW) aesthetics. The video is a masterclass in character consistency and tonal contrast. By placing an elderly woman in high-glamour, youth-centric environments—complete with a martini in hand and a younger man in her bed—the creator taps into a "subversive luxury" aesthetic that feels both aspirational and hilarious. The visual style is a polished "cinematic editorial portrait" mixed with "UGC vlog" vibes, utilizing warm, expensive lighting and sharp, high-fashion color grading.

What You’re Seeing

The video follows a day-in-the-life narrative. It starts with a close-up "vlog" style shot in a gold-toned luxury bedroom, transitions to a wide-angle bathroom shot, moves into the back of a high-end SUV, and culminates at a fashion show. The wardrobe is meticulously consistent: white silk pajamas with black piping, followed by a sophisticated black lace dress with matching cat-eye sunglasses.

Shot-by-Shot Breakdown

Time Range	Visual Content	Shot Language	Lighting & Tone	Viewer Intent
00:00–00:06	Darlene in bed, martini in hand, younger man sleeping behind her.	Close-up (CU), handheld vlog feel.	Warm gold, soft morning light.	The Hook: Shock value + character intro.
00:06–00:12	Darlene in a massive marble bathroom, gesturing with her drink.	Wide Shot (WS), static.	Bright, high-key luxury lighting.	Establish wealth and "larger than life" persona.
00:13–00:19	Inside a luxury SUV, Darlene looking out the window, then at camera.	Medium Close-up (MCU), car interior.	Natural side-lighting, high contrast.	Transition: Moving to the main event.
00:20–00:27	Red carpet interview with a reporter; paparazzi in background.	Medium Shot (MS), over-the-shoulder.	Flash-heavy, cool evening tones.	Social proof: She is a "celebrity."
00:28–00:34	Runway footage of models in oversized suits (KHAITE style).	Long Shot (LS), runway perspective.	Dramatic spotlighting, neutral palette.	Context: The actual fashion event.
00:35–00:45	Darlene backstage, "vlogging" her "bestie" closing the show.	CU to MS, handheld movement.	Industrial, fluorescent backstage light.	The Payoff: Insider access + final joke.

Why It Went Viral: The "Sassy Elite" Formula

The Subversive Archetype

This content succeeds because it subverts expectations of how an elderly person "should" behave. The "Sassy Grandma" is a proven viral trope, but elevating it to the level of a billionaire fashionista adds a layer of lifestyle envy. It captures the "Old Money" aesthetic but injects it with "New Money" attitude. The dialogue is sharp, using modern slang ("What's up hoes") which creates a hilarious juxtaposition with her sophisticated appearance.

Technical Excellence in AI

From a platform perspective, the watch time is driven by the sheer quality of the AI generation. Viewers stay to see if the AI "glitches," but it remains remarkably stable. The character's face, sunglasses, and even her jewelry remain consistent across six different environments. This level of technical polish signals "high-value content" to the algorithm, encouraging wider distribution.

5 Testable Viral Hypotheses

The "Contrast Hook": An elderly person saying something vulgar or modern in the first 2 seconds triggers an immediate "stop-the-scroll" reflex.
Brand Association: Mentioning real brands (KHAITE) or events (NYFW) makes the AI content feel grounded in reality, increasing shareability among fashion enthusiasts.
The "Martini Prop": Using a consistent, slightly "naughty" prop (the drink) helps define the character's rebellious nature.
UGC-to-Cinematic Pipeline: Mixing "vlog" shots with "professional" shots makes the character feel like a real person documenting their life.
The "Younger Partner" Trope: Including a younger, attractive background character creates a "mild controversy" or "aspiration" that sparks comments.

How to Recreate: From 0 to 1

Define Your "Hero" Character: Create a detailed persona. For Darlene, it’s "Wealthy, Sassy, 70s, White Hair, Cat-eye glasses." Use a consistent face-reference image.
Script for Contrast: Write dialogue that sounds like a 20-year-old influencer but is delivered by your character. Use "vlog" language like "You guys won't believe..."
Generate the "Anchor" Shot: Start with the bedroom shot. This sets the tone. Use prompts that specify "handheld phone camera" to get that UGC feel.
Maintain Wardrobe Consistency: Use a "Style Reference" (SREF) or specific wardrobe keywords (e.g., "black lace floral embroidery") in every prompt.
Environment Hopping: Generate 4-5 distinct scenes (Bedroom, Car, Event, Backstage). Ensure the lighting logic follows the scene (e.g., car interior vs. runway spotlights).
Voice Cloning: Use a tool like ElevenLabs. Choose a "Mature/Elegant" voice but adjust the stability to allow for "raspy" or "sassy" inflections.
Lip-Syncing: Use a tool like LivePortrait or Hedra to sync the generated audio to your character's mouth. Pay attention to the "martini" shots—ensure the glass doesn't clip through the face.
Edit with Pacing: Cut on the beat of the music. Use "punch-ins" (zooming in slightly) on funny lines to emphasize the humor.

Growth Playbook

Opening Hook Lines

"I woke up hungover as sh*t, but at least I look better than you."
"They told me I was too old for NYFW, so I bought the front row."
"My face is tighter than my pre-nup, and my drink is colder than my heart."

Caption Templates

The "Insider" Template:
NYFW Day 1 and I’m already over it. 🍸 [Value: Aesthetic visuals] Who else is surviving on martinis and spite this week? [Engagement] Tag your fashion bestie. #NYFW #AIDesign #DarleneCore

Hashtag Strategy

Broad: #AI #Fashion #NewYork #Vlog (Reaches general audiences)
Mid-Tier: #AIVideo #DigitalFashion #OldMoneyAesthetic (Reaches tech/style enthusiasts)
Niche: #Khaite #FashionHumor #SassyGrandma (Reaches specific subcultures)

FAQ

How do I keep the face looking the same in every shot?

Use a consistent "Face Reference" image and include specific descriptors like "white wavy bob" and "black cat-eye sunglasses" in every prompt.

What AI video tool is best for this?

Runway Gen-3 Alpha or Luma Dream Machine are currently best for maintaining character consistency across complex movements.

How do I avoid the "AI look"?

Add "film grain," "slight motion blur," and avoid "perfect" lighting; use "handheld camera" prompts for a more natural feel.

Is this better for TikTok or Instagram?

Instagram Reels generally performs better for high-end "aesthetic" and "fashion" AI content due to its visual-first culture.