Work we have delivered, grouped the way the studio actually produces it.
These are a handful of examples. The full archive runs many times deeper — tell us what you are training and we will show you the work that fits it.
Re-photographing real objects and real people to a fixed specification — a new angle, a new setting, a new outfit — so a model sees the same subject under conditions it was never originally captured in. This is where the contributor network matters most: the variation comes from hundreds of different rooms, cameras and lighting conditions rather than from one studio repeating itself.
Our widest reshoot by headcount: more than two hundred contributors working in parallel, each covering objects and people in their own interiors and their own light. That spread is the point — five thousand sets assembled from two hundred different rooms carry variation a single studio cannot fake — and the two-week turnaround held because briefing, review and delivery all ran through one pipeline.
5,000 sets2 weeks200+ crew
A fixed-composition brief: the same tool, a new surface, a centred subject flanked by props. Repeating one layout three thousand times across a hundred shooters is a consistency problem rather than a photographic one, so the framing, the crop and the prop rules were settled before anyone shot, and every set was checked against them on the way out.
3,000 sets4 weeks100+ crew
Re-photographing one object from angles it had never been shot from, the top-down view of the stump included. Multi-view coverage of a single subject is what teaches a model that two very different images are the same thing, and a thousand such pairs came back inside two weeks.
1,000 sets2 weeks100+ crew
Garment transfer done for real — a reference outfit, a live model, a photograph rather than a render. Three hundred contributors gave the set its range of faces, body types and lighting, which is precisely what a virtual try-on model needs and precisely what a studio cast of ten cannot supply.
2,000 sets4 weeks300+ crew
Motion held at a fixed camera relationship: subject running, eyeline to camera, side angle throughout. Directing nine hundred sets of a moving person to a single framing spec is a briefing discipline, and it is the part of crowd-sourced capture that most often goes wrong.
900 sets2 weeks100+ crew
The most tightly specified brief in this group — two named subjects, plain beige studio background, a hugging pose, both looking to camera. Small volume, no tolerance: two hundred sets in which every variable except the people was held constant.
200 sets2 weeks50+ crew
Generation work where the binding constraint is identity — the image has to change while the person, the product or the scene stays recognisably itself. We run it as a loop: generate, review, reject, regenerate. That is why the smallest teams in this portfolio sit in this group, and the highest weekly output does too.
Identity-preserving composition: the character out of the source photograph, now holding an object that was never in it. Five people turned out five thousand sets in a single week, which is the difference a tuned generate-and-review loop makes against ad-hoc prompting.
5,000 sets1 week5 crew
The hardest generation brief we have run. Three separate source images had to become one courtyard scene with the people interacting naturally, while every face, body type and garment — and the original camera angle, lighting and composition — survived the composite intact. Ten specialists, six thousand sets, two weeks; the volume came out of the review pass, not the prompt.
6,000 sets2 weeks10 crew
A three-way composite: wardrobe from one image, location from another, a person placed convincingly into both. Underground parking is an unforgiving test — hard overhead light, strong reflections, shallow depth — and three thousand sets cleared it in a week with a team of five.
3,000 sets1 week5 crew
Our highest single-week output: ten thousand advertising posters generated from source product imagery by a team of ten. At that cadence the work is pipeline design — templating, batching, automated rejection — far more than it is image-making.
10,000 sets1 week10 crew
Moving image, from single-object cinematography to synchronised two-camera sessions and a forty-thousand-clip sports library. Video is where scheduling becomes the hard part, and where running recruitment, shooting and review as one continuous line rather than as three departments pays for itself.
Object cinematography rather than stills: controlled moves around a product, consistent lighting, footage cut to a fixed length. Two hundred sets in a week from ten shooters, each frame built to hold up when a model samples it one at a time.
200 sets1 week10 crew
An editing-only engagement — source footage in, finished cuts out, five hundred of them in a week from a team of seven. Editing at that rate needs a template that survives contact with mismatched source material, and that is where most of the work actually went.
500 sets1 week7 crew
Point-of-view footage in the native vertical format that social platforms, and the models trained on them, expect. Two hundred-plus contributors shot on their own phones in their own surroundings, so these eight hundred sets carry genuine device variation instead of one studio camera's signature.
800 sets3 weeks200+ crew
Long-form multi-camera capture: five minutes of continuous training footage from two synchronised angles, a hundred times over. Running two hundred people through a protocol that strict for three weeks is a coordination problem before it is a filming one, and it is why this kind of dataset is rarely available off the shelf.
100 sets3 weeks200 crew
The largest single delivery in this portfolio — forty thousand one-minute sports videos across eight weeks with a standing crew of two hundred. Sustaining roughly five thousand finished clips a week for two months only works when recruitment, shooting, review and hand-off run as one continuous line rather than as four sequential stages.
40,000 sets8 weeks200 crew
A deliberately diverse cast: children of different nationalities, filmed to one specification. Demographic breadth is the entire value of this set, and it came from more than a hundred contributors recruited across our network rather than from whoever happened to be available locally.
300 sets3 weeks100+ crew
Small, technical and post-heavy — live action shot to be extended with visual effects. Thirty sets in two weeks with five people, deliberately sized so a specialist crew could stay on every single frame from plate to composite.
30 sets2 weeks5 crew
Two- and three-dimensional production: renders built from UV textures, floor plans paired with models that match them exactly, and hand-drawn game illustration. Part of this group is high-volume production engineering and part of it is careful drawing — the distance between the largest and the smallest project here is the widest on the page.
Forty thousand 3D renders built from UV textures, each showing the half of the object its source texture describes. Texture-to-geometry work at this volume is production engineering: thirty artists on a shared rig and asset pipeline, four weeks, one output spec that never moved.
40,000 sets4 weeks30 crew
Paired 2D and 3D output of the same apartment, and the two had to agree exactly — an accurate floor plan and a dimensional model matching it measurement for measurement. That correspondence is the dataset. A thousand matched pairs came from twenty artists over three weeks.
1,000 sets3 weeks20 crew
Animated game illustration — the smallest volume here and by far the most hand-made. A hundred and fifty pieces over four weeks from a team of five, where the value sits in the drawing rather than in the throughput.
150 pieces4 weeks5 crew
Post-production and design — restoration, reference-driven composition, vector illustration, infographics and interface layout. All of it is judged on consistency rather than on any single image, which is why the review stage in this group carries as much weight as the production stage.
Information design: seven hundred and fifty infographics built to a single visual system in three weeks. Ten designers working from one shared template library is what keeps a set this size legible as a set, instead of as seven hundred and fifty unrelated layouts.
750 sets3 weeks10 crew
Restoration and modernisation in one pass — physical damage removed, then a contemporary digital version of the same photograph. Every one of the four thousand sets is a before-and-after pair, which is what makes it trainable rather than merely a retouching job well done.
4,000 sets3 weeks30 crew
Reference-driven Photoshop composition at volume: five thousand sets in four weeks from thirty retouchers. The brief was open by design, so the control came from the review stage instead — one shared quality bar applied to every set before it was allowed out.
5,000 sets4 weeks30 crew
Vector work in Illustrator, built to supplied references. Six hundred sets over three weeks from ten illustrators, delivered as clean resolution-independent artwork rather than as raster approximations of it.
600 sets3 weeks10 crew
A second Illustrator engagement, larger and looser than the first: a thousand sets in three weeks with fifteen illustrators. Vector output scales in a way raster work does not, which is what let this set grow without a matching growth in review time.
1,000 sets3 weeks15 crew
Interface and layout design across Figma and Photoshop — three thousand sets in four weeks from twenty designers. Mixing a design tool with a raster editor in one pipeline is a file-handling problem before it is a design one, so the hand-off rules were fixed before the first set was made.
3,000 sets4 weeks20 designers
Engagements we own from the first sketch to the final photograph: design, physical production, then shooting the finished object. Nothing waits on an outside supplier between stages, and that is what keeps a chain this long inside a four-week schedule.
The full chain inside one engagement: design created from scratch, produced in print, then photographed as a physical object. Ten designers and five photographers delivered three thousand sets in four weeks, and owning every stage is what makes that schedule possible — nothing waits on an outside printer or a separate photo crew in between.
3,000 sets4 weeks15 designers & photographers