Rudyard Cretenet has spent many years in VFX as a compositor and 2D Supervisor, on live-action shots that mix CG and plates, with Nuke as his main tool. Before cohort 1 of our Generative AI Bootcamp for Film and TV, he had never opened a generative AI tool.
He did not build one project. He built three, to learn what these models do to an image and to a shot: architecture studies for a Paris that never was, test shots for a film before the shoot, and a one-minute film. He presented them to the final jury on 11 September 2026.
🎞️ Goosebumps, a one-minute film
The short above is the film, as he showed it to the jury. The premise: between humans, fear, admiration or inspiration need no explanation. You say you are afraid, and the other person knows. Try to explain the same thing to a machine, and it may be impossible. Goosebumps imagines the journey of an AI trying to learn an emotion, and the film was itself generated by an AI. In its own words, strong emotions trigger adrenaline, adrenaline stimulates the nervous system, and that is when you get goosebumps.
The images and the voice-over are generated with LTX 2.5 in ComfyUI. Rudyard Cretenet worked in groups of three shots rather than shot by shot, and fed each generation back into treatment passes until the look held. The edit was done in DaVinci Resolve. The music is Celestial Calling (Original Mix) by Fog Temple, licensed through Slipstream Music.
🏛️ A Paris that never was
The first project is an idea Rudyard Cretenet has carried for years. Take the footprint of a Haussmann building in today's Paris, keep the base on its square, and merge it with the iron and glass of the 19th-century train stations, Gare Saint-Lazare and its kind, with art nouveau and elongated windows. The images were made in ComfyUI with Z-Image Turbo and Qwen Image Edit 2511. He paints where he wants the frame to change, generates, and keeps one result out of many prompts and retakes. Then he turns the prompt into a recipe and refines it, quality and lighting, until it holds across variations. In Photoshop, he says, the warping and the lighting matches alone would have taken far longer.
Study 1: a basilica, art nouveau merged with an iron-and-glass structure, and a pinch of art deco in the shape.
Study 2: a tower placed behind a first building, with a facade like a station roof.
Study 3: the prompt turned into a recipe and refined. Twin towers, variations in the same style.
🎬 Test shots before the shoot
The second project is pre-production. Rudyard Cretenet has worked on films where you cannot go to the set, or to the country, and where the director still needs test shots. The example is a film project set in Belgium. After many generations he had a start frame: a woman who looks like his actress, a man who looks like his actor, the right perspective and lighting, and a city that was wrong.
From there, everything is Qwen Image Edit 2511. The background becomes a Belgian street (the model knows cities and countries, so the prompt can name them). The frame is extended. A building he did not like is replaced, another sky is tried, then the previous sky is brought back while part of the new frame is kept. He changes the camera axis by a few percent and draws where the change should apply, moves the horizon line, changes the focal length by typing it, adds cars, and generates a clean plate that takes a few tries and a little paint.
One test shot, many versions with Qwen Image Edit 2511: background iterations, building and sky replacement, change of camera axis, clean plate.
🎓 Built in the Bootcamp
Rudyard Cretenet came to cohort 1 with a compositor's eye and no experience of generative AI. He left with three pieces of work and a new set of tools in his kit: generative AI in ComfyUI now sits next to Photoshop and Nuke.
Cohort 1 of the Generative AI Bootcamp for Film and TV finished on 11 September 2026 with a group of six participants, each presenting their own project, built in ComfyUI across the sessions, with a midpoint review and a final review by a jury of working AI professionals.
Cohorts 2 (from 2 October 2026) and 3 (from 6 January 2027) are open for enrollment. You bring one production case, and you build it. Learn more about the Bootcamp.
🔗 Sources & Links
🎬 The project:
- Goosebumps, the short on YouTube
- The announcement on LinkedIn
- Celestial Calling (Original Mix) by Fog Temple, on Slipstream Music