/

/

ROSA

Haunted by silent memories, a woman traverses surreal landscapes of concrete and crystal, searching for a path through the architecture of her own past.

Concept

Produced for the OpenArt Music Video Awards, “Rosa” explores the architecture of memory. The film navigates three contrasting visual worlds—Brutalism, Crystal, and Museum—acting as emotional mirrors for the protagonist. It is a cinematic meditation on navigating an exit from one’s own past to reclaim a grounded sense of reality.

Execution

The production utilized a multi-shot lip-sync strategy, generating three angles per vocal act to ensure dynamic pacing. Character consistency was achieved via face-swapping and reinforced through visual markers like a signature yellow hair streak. Inpainting was used to polish compositions, while AI-generated SFX provided tactile realism.

Director’s Note

“Rosa” taught me to use the environment as a narrator. By leveraging specific keyword triggers, I maintained thematic unity across various models. This project proves that even when using disparate generative tools, a strong creative direction can weave them into a unified, high-fidelity narrative experience.

Release Date

November 14, 2025

Runtime

00:02:00

Directed by

Taruma Sakti

Technique

txt2img, img2img, img2vid

AI Platform

OpenArt (Image, Video, Lip-sync, Upscale), fal.ai (SFX), Higgsfield AI (Image), Google AI Studio (LLM)

Text/LLM Model

Gemini 2.5 Pro, Gemini 2.5 Flash

Image Model

Seedream 4.0

Image Editing Model

Seedream 4.0, Higgsfield Face/Character Swap

Video Model

Kling 2.5 Turbo Pro, Kling 2.1 Pro, Seedance 1.0 Pro

Lip-sync Model

Hedra (Fast)

Upscale Model

Topaz Labs

SFX Model

Mirelo AI SFX v1.5

Other Tools

Shotbuddy v3 (AI Asset Manager), Typingmind (LLM Interface), Adobe Premiere Pro (Video Editor)

Music

Hello Yellow by Zorro, Lupus Nocte

posted by

Recent Works