The AI plate and the live subject are both describing motion, in different ways. Three AI-generated environments and one real, photographer-captured plate, shot on the same VP stage, with the model's movement captured live via shutter drag.
Unlike the shared workflow module used elsewhere in this set, this shoot ran without Marco. The photographer's own laptop did both jobs, live.
No Marco on this shoot. One laptop only, the photographer's own: MacBook Pro M2 Max 16″ (96GB Unified Memory / 4TB SSD Storage).
The same laptop ran both sides live: Capture One tethered for stills, and Photoshop for live plate playback, sent out through the ROE screen as an extended display. Five plates shot across the day, four AI-generated and one real.
Placeholder figures, to confirm with Doma: Midjourney as the primary AI tool, around 8 hours of work across the AI plates, each iterated twice before the final version.
Shutter drag is a deliberately slow shutter speed that records motion as controlled blur instead of a sharp, frozen frame. It's the term photographers actually use on set, and it sits inside the wider category of intentional motion blur (IMB), any deliberate use of blur as a choice rather than a mistake.
For an elevated ecom campaign, what matters is that the backdrop can move too. Three of the four plates here are AI-generated, each with its own built-in sense of drift and direction, tuned live to match the model's shutter-drag movement so background and subject read as one coherent shot, temporal coherence. That's the flex a VP stage offers a brand: a location redirected in real time to match whatever story the product needs, no location shoot required.
All figures above are read directly from the camera metadata embedded in the delivered files, not from memory or estimation. Shutter speed is not uniform across the whole shoot: Fabric Flow, Cube Room and Side Street were all captured at 1/30s, while the Plaster Texture Recess set used a wider range (1/13s and 1/40s recorded across its frames).
The one plate in this shoot that is not AI-generated: a real location, shot by Helen Labourt on a Canon EOS R5, used as the VP background for this look.








AI-generated environment, edited in real time by a remote AI artist during the shoot. This section is about three things: the scale of the fabric itself, the colour changes made live, and how fast the LED backdrop updated in response.
The drape here is the clearest example of temporal coherence in the whole shoot. Its motion supplies velocity cues, oversized enough to read as speed and direction even before the eye reaches the model, and the AI artist could shift those cues live (colour, drape angle, density) to stay coherent with the model's own shutter-drag movement on each pass.



This set underpins the whole real-time argument for AI on a VP stage: skin tone, make-up and garment all had to be re-read by the AI artist as the plate colour shifted live, in the same session, against the same model, without a reshoot. Close-up BTS of make-up and garment coming from Helen to make this point explicit.






Every environment in this section was built by Casey J Bird, our AI artist. Two things still need confirming before this goes live.
As this shoot leans on AI-generated environments, it's worth a short note on where that sits from an IP standpoint, current as of the shoot date.
[Placeholder for Kelsey's own words, a light-touch comment on ensuring AI-generated imagery isn't trained on other people's copyrighted work.]
A gallery of looping clips from the day: live tethered capture on the volume, and a pan across the set floor and screen. The parquet floor was embraced rather than disguised, one of the standard approaches to the VP flooring problem.



AI-generated environment: a rust-red textured recess, distinct from the Fabric Flow set's drapes, and shot at a different shutter range (1/13s and 1/40s recorded across its frames).





Cost note: no physical set was built for this environment. The textured recess exists only on screen, so there's no plaster, framing or dressing to construct, transport or strike.
AI-generated environment. The story here is directional light: hard, raking stripes across the walls and floor, motivated by an unseen window just out of frame.



Cost note: same principle here. No cube, no window, no raking-light rig was physically built; the architecture and directional light are entirely generated, so the set-build cost for this look is zero.
This shoot was published in Artego, a fashion and glamour platform known for showcasing emerging photographers with professionally shot editorials, reaching an audience of 128,000+ across print and digital.





Placeholder quote below shows the tone and layout this section will use, swapped for Doma's real words and headshot once received.
Example only, something like: "What I loved about shooting this as VP was chasing fluid motion in-camera, letting the fabric blur into the frame rather than faking it afterwards."
Confirmed team and Instagram credits for this shoot. Doma's quote and headshot, plus more BTS, still to follow.
Helen Labourt: Side Street plate capture (Canon EOS R5, Bonnieux, France).
The same five lenses used across every case study in this set, each one already carried by a section above. This is the recap.
Shutter drag and temporal coherence, reframed for elevated ecom: a backdrop that moves live with the model, no location shoot required.
No physical set was built for either AI environment. The textured recess and the cube room exist only on screen, at zero set-build cost.
One real location, Bonnieux, France, shot once and reused as a VP background rather than travelled to and rebuilt for every campaign.
No AI-generated imagery is presented as real. Casey's exact process and a light-touch IP note from Kelsey are both still pending confirmation.
This shoot ran without Marco. One laptop, the photographer's own, carried both live plate playback and tethered capture end to end.
A working draft: send feedback on the Fabric Flow colour story, the Plaster Recess process sequence, and the quote above, and this becomes final.


