Adobe’s Project Doki turns written stories into editable videos

Adobe Research’s Project Doki is an experimental video creation tool that lets users build and refine generative videos through freeform writing. Its text-native interface keeps characters, settings, visual style, story development and audio in a single document. The research received a Best Paper Honorable Mention Award at UIST 2026, the ACM Symposium on User Interface Software and Technology.

Users can adjust in-line visual previews and work with a conversational AI agent as they develop their stories. Once a story is ready, the tool compiles the text into video shots. Users can then refine the results through text edits or the agent, giving them a document-based alternative for controlling generative video.

The research team recruited ten people with varied video experience to test the tool over a week. Participants rated the experience as excellent on the System Usability Scale and reported faster progress from ideas to output, alongside a better understanding of narrative flow. Novices valued being able to create visual stories they previously could not, while experienced users highlighted improvements to ideation and storyboarding.