Google’s Fabula demo

Story-writing software usually predicts the next sentence. Google Research’s Fabula instead helps writers reshape plot, scene order, and story beats through back-and-forth editing. (research.google) Google scheduled a Fabula demo for Wednesday, April 15, at the Association for Computing Machinery Conference on Human Factors in Computing Systems, or CHI 2026, in Barcelona. The conference site describes CHI as a leading international meeting on human-computer interaction. (research.google, chi2026.acm.org) Google Research said Fabula was co-designed with 42 expert writers. The demo description says the tool is built to help authors “structure & refine stories” and uses what Google calls “convergent iteration” to support creativity. (research.google) In plain terms, that means the system is aimed at story planning before draft polish: writers can adjust the outline, regenerate parts at different levels, and keep revising the narrative shape instead of accepting one long machine-written draft. A Google-linked description of the project says Fabula works across scenes and beats, the smaller units inside a scene. (research.google, lilting.ch) That design puts control on the writer’s side at a moment when many artificial intelligence writing products market speed and full-text generation. Fabula is being presented as a research prototype at an academic conference, not as a public product launch. (research.google, lilting.ch) The project also fits Google DeepMind’s earlier work on assisted scriptwriting. Its 2022 Dramatron paper described a co-writing system for screenplays and theater scripts and reported a user study with 15 theater and film professionals. (arxiv.org, deepmind.google) Google’s public Dramatron materials also said the tool “was not conceived or evaluated to be used autonomously.” Fabula appears to continue that human-in-the-loop approach, with more emphasis on planning and revision workflows than on one-shot generation. (github.com, research.google) Outside Google, researchers have kept reporting the same problem with machine-written fiction: long stories often lose coherence, control, or narrative consistency. Recent academic work on story generation and narrative interfaces has framed structure, evaluation, and user agency as open problems rather than solved ones. (arxiv.org, dl.acm.org, arxiv.org) Fabula’s demo lands in that gap between autocomplete and authorship. Google is showing a writing partner for rearranging the skeleton of a story, not a machine that claims to finish the book. (research.google, github.com)