What’s up everyone. I leveled up to 37 yesterday and decided to run a fun experiment with my faceless marketing strategy.
Instead of doing the standard "Happy birthday to me" selfie, I built a hyper-realistic 1993-1997 era R&B/Talk Show set and let my two custom AI virtual personalities (Mizz Lexi & Kayla Kriss) handle the heavy lifting. The goal? Have them wish me a happy birthday and smoothly convince the audience to drop some birthday tips in the CashApp.
The Stack & Workflow:
Reference Locking: Used Nano Banana to lock in the exact facial structures, wardrobe (the satin bomber vs. the gold wrap dress), and the neon Chill Zone Studio background.
Motion Control: Ran the generated stills through Seeddance 2.0. Used very specific, slow camera movements (like a slow, sensual dolly-in for the second character) to keep the 90s music video vibe authentic without warping the facial geometry.
Audio Mix & UI: Scripted precise pauses and chuckles for the Text-to-Speech to sound like a live broadcast, layered over a vinyl crackle. I also slapped the CashApp QR code right on the screen to make the "tip jar" completely frictionless.
I'm heading to the Chicago AI Film Festival today with this proof-of-concept in my pocket, but I value the critical eyes in this group.
Feedback Request: How are you guys handling the transition cuts between two different AI character generations in the same scene? I used a hard cut here to switch between the girls, but I'm thinking a subtle audio swell or crossfade might glue the scene together better.
Would love to hear your critiques on the lighting, motion, and the overall execution!
Have a great time at the Chicago AI Film Festival today, Ern. Show them exactly what the Chill Zone Studio is capable of.