silentfleetkk/vellum-reel

VellumReel is a local video production system that turns text, such as book notes, reading reflections, or complete stories, into vertical 9:16 narrative videos. It does not try to be a general purpose video editor. Instead it focuses on the parts of text-driven video that are easy to get wrong: splitting narration into segments, Chinese subtitles, narrative pacing, camera movement, keeping timing in sync, rendering many videos in batch, and automatically checking the finished files. It supports two modes: one for explaining a book, and one for novels, podcast style stories, or full chapters. Every project, its subtitles, camera shots, and story beats are defined as reviewable JSON files rather than hidden settings. It has specific support for Chinese text, including splitting sentences naturally, merging pauses, emphasizing keywords, and handling both dialogue and quotation style subtitles. The length of each shot is driven by the actual recorded narration audio, and the audio itself can be stitched together from segments and normalized to a consistent loudness level. Camera work is deliberately restrained rather than flashy, with per shot motion, intensity, color grading, and transitions, plus a consistent visual frame and reading progress indicator across the whole video. Before a full render, the system checks that the footage, frame size, and timeline are all valid, and after export it uses FFmpeg and ffprobe to automatically verify the result. To get started you need Node.js 20 or newer along with FFmpeg and ffprobe installed, then you install the project's dependencies and can try two included example demos, a short book explainer and a short original story, before copying an example into your own private folder to produce something with your own content. Private project folders and media files are excluded from the repository by default. The project's own code is released under the MIT license, but it depends on Remotion, which has its own separate license that is generally free for individuals and small teams but may require a paid license for larger organizations or commercial automated use.