A Mansion Without Doors

Music video for Fabio Bernardi

This project began as a commission but soon became an exercise in world-building. A closed, symbolic, unsettling world. A place where rules aren’t explained, but felt. This isn’t a music video in the traditional sense — it’s a system of images that revolve around symbolic tension. The setting is a witch mansion, not in the kitsch or fantasy sense, but as a psychic architecture: ritualistic, feminine, seductive, structured. No magic clichés, no glowing effects. I wanted a dark, polished aesthetic, where tension is constant but never tips into horror or overused fantasy tropes.

Imagery and Symbolic Framework

This world is built on a visual language of symbols:
Black crows as silent aides to the headmistress.
Moths as fragile agents of transformation.
The white crow as anomaly, a sacred glitch.
The forest as the outer edge — a threshold, an escape that must be earned.
And the headmistress as perfect captivity: power that no one wants to leave.

The narrative moves quietly. The transformation belongs to the male protagonist, but it’s triggered by her — the rebel. There is no dialogue, no exposition. Only attraction, only disruption.

Visual Hierarchy and Coherent Codes

To create inner structure, I designed a visual hierarchy based on hair length.
Shaved heads mark the initiates.
Long hair belongs to those at the beginning or the top.
This system works symbolically (discipline, surrender, status) but also technically: it’s far more stable and consistent for generative AI than clothing, uniforms, or detailed costume design. A clear sign that cuts through.

Aesthetic: Slowness and Suspension

The visual language is slow, ritualistic.
Scenes aren’t narrative — they’re atmospheric.
Animations are minimal, like held breath. Each frame is treated as a living tableau. I avoided spectacle in favor of dreamlike stillness, where discomfort and elegance coexist. Emotions aren’t described — they’re embodied in gesture, posture, light. The beauty here doesn’t soothe. It unsettles.

Technology and Process

Images were generated with Midjourney V6, using carefully crafted prompts to avoid any fantasy or pop aesthetics. In certain cases, I deliberately left atmospheric keywords in the prompt (like dark garden in the underwater scene), even when not visually present — to preserve narrative tone consistency between unrelated images. In this workflow, AI is not a creator. It’s a medium that responds — if led with clarity and precision.

I used Photoshop where needed to stabilize character design and fix inconsistencies between sequences. All video outputs from Midjourney were visually strong but low in resolution. For that reason, I upscaled all animations using Topaz Video AI, achieving a final quality fit for professional use (editing, projection, publication). I tested Runway (visually unstable for cinematic language) and Kling (visually promising, but with unclear licensing conditions).

Compositional, not Illustrative

I didn’t want to illustrate the music. I wanted to give it a visual counterpart — something that didn’t simplify it, but expanded its field. This project explores how generative AI, when used with intention and control, becomes a compositional tool, not a shortcut. A way of building atmospheres, not decorating scenes. The result is a closed system — dense, stylized — that moves between fashion, symbolism, and dream cinema. A crack in the surface of imagery.

Tools used:
Midjourney V6 – Photoshop – Topaz Video AI – Premiere – AI Color Grading (manual fine-tuning)