A full DTC skincare ad generated end to end: wake, bathroom, open the serum, apply, result, product hero on white. 16:9, hook inside the first second, sound designed and levelled for social. Built to the shape a skincare ad actually needs — problem, use, result, product — rather than assembled from whatever generated well.

Link:

A 21-second beauty commercial built end to end: brand, product, script, voice direction, edit. Spec work — the brand is invented, there is no client. The cut is built to the voiceover rather than the reverse: gaps between spoken lines are measured and each shot sized to the line over it, so the brand name lands on the bottle reveal and the lather lands on “from the very first wash.”

Link:

Four opening variants for the same streetwear brief — a night street, Times Square, a graffiti alley, a rooftop — with the same two characters holding across all four. Self-initiated. This is what testing four openings against one brief actually looks like, rather than one execution presented as the only idea.

Link:

A single repeatable ad device executed across ten different worlds: the product becomes its own environment. A watch face opens into a marble hall, rings shatter into a showroom, a book erupts into the library. One idea that survives ten different briefs. Produced from supplied briefs.

Link:

Launch-event invitation campaign for a streetwear label. A red envelope carries the invitation across four unconnected worlds, each scene resolving by the frame flooding red. Paid client work at Hybriddot, who held the client relationship — script by their writers, visuals mine: generation, shot construction, take selection.

Link:

One AI presenter holding face, voice and character across five separate environments — private jet, boardroom, penthouse at night, garage — with dialogue intact throughout. Holding a single character across a long sequence is where AI video usually breaks.

Link: