Everyone on a project has ideas for a shot. Fewer people can draw them. Julien Chateau built a tool that makes a rough sketch enough: you draw where the characters and the camera go, you add one line of text, and you get back a clean storyboard frame that keeps your framing and your poses. Artists who are not comfortable with drawing get a board they can share with the team. Artists who draw well iterate faster, with the same quality on every frame.
Julien Chateau, CG Instructor, built it during cohort 1 of our Generative AI Bootcamp for Film and TV and presented it to the final jury on 11 September 2026.
🎬 Who it is for
Julien has worked for about twenty years in commercials, films, animation and trailers, and he teaches CG. His students have to deliver storyboards, and some of them draw less well than others. The storyboard tools he tested start from a text prompt: you describe the story and the software invents the frames. His tool starts from the drawing. Nothing is generated until someone has sketched the shot, on paper, in Photoshop or in the app itself. The sketch sets the framing and the poses. The model cleans up the drawing and keeps everything in place.
🖥️ The tool
It runs in the browser. A project holds sequences, a sequence holds shots, and each shot has a number, a camera, an action and a duration. You can reorder shots, and they keep their information. You can link two frames into one shot, with a start frame and an end frame. You can import a scan or a Photoshop file, draw in the app, zoom in and out, and draw over a generated frame to correct it.
Every generation is a ComfyUI workflow. Julien exported his workflows as JSON files and calls them from the app, which talks to a ComfyUI server. He wrote the app during the cohort.
✏️ From sketch to frame
The main step uses Qwen Image Edit, an image editing model, with a LoRA Julien trained during the cohort. A LoRA is a small add-on trained on a few examples that teaches a model a style or a task. His dataset has 44 examples: for each one, a finished drawing and the rough sketch. The LoRA has a trigger word, and it gives every generated frame the same drawing style. The artist adds a short description of what is in the sketch, and the result keeps the composition and the poses exactly where they were drawn. Julien tested several models, Z-Image among them, and kept Qwen: it respected the framing and the poses best.
🧰 Editing tools
The same workflows power the editing tools:
- Adapt to ratio. A frame drawn in the wrong format is completed to the project's ratio. The model paints the missing part (outpainting).
- Upscale. A small or low-quality input becomes a sharper frame.
- Character swap. Each character has a name, a short description and a model sheet (a reference drawing). Click a character in the frame, pick another one from the list, and the model replaces it. The same tool corrects a character the sketch-to-frame step did not draw right.
- Relight. One line of text, for example golden-hour light through a window and a fireplace lighting the floor, and the frame is relit.
- Extend the shot, and draw over anything the model missed.
🔭 What comes next
Julien keeps developing the tool for his own work, and he plans to publish it once it is complete, so that anyone who can picture a shot can storyboard it.
🎓 Built in the Bootcamp
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. Julien's storyboard tool is one of them.
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.