m4n0.ai.archive is a standing collection of AI-generated images and videos made from 2023 to the present. It runs alongside the commissioned work, not after it: the archive is where models get tested, prompts get pushed, and the visual language behind the production cases takes shape.

The pieces are not one look. Engravings, creatures, plants, cityscapes and portraits sit next to each other, each tagged and dated, filterable by series, media type and time. Read as a whole, it tracks how the tools changed across four years, and how the practice changed with them.

Browse the full archive: m4n0.ai.archive

Audiovisual Production, Spatial Tech Architecture & Creative Collaboration · with Gracias Grecia

Tecnologia [in]visible is an exhibition by Mobile World Capital Barcelona and the Generalitat de Catalunya at Palau Robert, curated by Genís Roca. It traces twenty years of technological change through a three-act journey: technology as object, as system, and as something we shape collectively. On view 12 June to 27 September 2026.

Gracias Grecia handled exhibition design and creative direction. I collaborated on the creative and led the audiovisual production: generative video, a six-channel testimonial installation, subtitling, and the delivery schedule.

Generative video
Creative direction and full video production for an interactive room where visitors reveal footage by moving a polarizing filter. Eight everyday-technology use cases appear in two synchronized states: a photorealistic scene of the technology in use, and its X-ray counterpart, generated from a shared seed. The pipeline pairs realism and infrastructure, making invisible systems legible.

Six-channel testimonial installation
A six-channel multiscreen piece built from real testimonies about technology, produced end to end including the shoot. Before filming, a fully AI-generated reference version, with synthetic people, scripted statements, and cloned voices, previsualized the piece and set framing, pacing, and channel logic across all six screens. The final work is the filmed testimonies, synchronized across the six channels in two editing structures, one dialogic and one contemplative.

Stack: ComfyUI, Runway, Nano Banana, Claude, ElevenLabs, Whisper, After Effects, Premiere, HTML/JS

Poster art direction and full production for the theather play Sarmiento, La Clase. Restored and color-graded historical portraits of Domingo F. Sarmiento form the base, onto which a faceswap maps actor Juan Leyrado in character. Upscaling then prepares the composite for photographic realism at large-format print size.

A custom AI pipeline combines the historical portrait with the actor’s features through a controlled faceswap, correcting distortions by hand rather than leaving them to automation, and finishes with high-fidelity textures and a chemical-emulsion look for large-format urban print.

Technical Process:
Restoration & Color: Neural texture recovery and 19th-century chromatic colorization.
Hybridization: Feature exchange via ComfyUI nodes, balancing historical records and actor references.
Refinement: Manual gaze adjustment and high-fidelity upscaling for outdoor printing.

Cover art and motion for a Spotify release built around the look of a popular anime. The brief called for the reference series’ unmistakable feel while staying clear of third-party IP, so the workflow abstracts style rather than reproducing characters. Generative iterations settle on an original image that reads as fan-adjacent without infringing.

Production Meeting is a set of wax figures cast as candles: women in period dress seated at laptops, lit during the exhibition and left to burn down over its run. The scene rewrites historical labour as digital labour, and the material carries the argument by consuming itself while the audience watches.

The figures are not modeled by hand. A flat black-and-white engraving-style image is generated to prompt, converted to a depth map through a ComfyUI workflow, then displaced onto a plane in Blender to build the relief. That relief is 3D printed as a master, a negative mold is taken from it, and the final pieces are cast in pigmented wax. Generation sets the image; the mold and the flame finish it.

Technical process:
1. Image: flat engraving-style image generated to prompt.

2. Depth: ComfyUI workflow converting the image into a depth map.

3. Relief: depth map displacing and decimating a plane in Blender.

4. Fabrication: 3D-printed master, negative mold, cast in pigmented wax.

Exhibition: figures lit as candles, self-consuming over the run of the show.

AI Tools: Midjourney, ComfyUI, Blender

Acquired by the Belvedere’s Bildarchiv & Ursula Blickle Video Archiv.

A video artwork on how we draw the line between natural and artificial. It centers on a recorded dialogue with an artificial agent whose answers, question by question, are pushed toward an absurd conclusion, staging the moment the category of “nature” collapses under its own definition.

The visuals are built as a chain of generated keyframes, not filmed footage. A reference set of video-game textures and 3D assets sets the starting point, and from it still frames are generated to follow each beat of the script. Runway animates between them: three frames per segment, moved through as a slow overhead flight, with the last frame of each export carried in as the first of the next. The result is one continuous descent through synthetic landscapes that never resolve into a real place.

Technical process:
1. Reference: Set of video-game textures and 3D assets.

2. Frames: stills generated to match the script, beat by beat.

3. Motion: Runway interpolating three frames per segment as an overhead flight, chained by reusing each segment’s final frame as the next one’s first.