vox-director: an automated pipeline for creating editorial paper-collage explainer videos from a single topic

What it solves

Vox Director automates the creation of high-quality, editorial-style "paper-collage" explainer or advertisement videos. It eliminates the manual effort of scripting, designing keyframes, animating, and adding audio, allowing users to turn a single topic into a finished video file.

How it works

The project operates as an agent skill that integrates with coding agents (like Claude Code or Codex) and the Atlas Cloud API. It follows a structured pipeline driven by a beats.json file:

  1. Beat Mapping: Generates a narrative arc and a set of beats for user approval.
  2. Style Bake-off: Renders a few theme options for the user to choose the visual look.
  3. Keyframe Generation: Creates a rich collage poster for each beat using text-to-image models.
  4. Motion: Animates the posters using image-to-video models or a local keyframe engine for precise, piece-by-piece assembly.
  5. Audio: Generates narration via TTS and background music.
  6. Assembly: Uses ffmpeg to combine all elements, duck music under voice-overs, and burn in captions.

It supports three input modes: B-roll (topic-to-video), A-roll (re-styling existing talking-head videos), and C-roll (anchoring a single still photo into the collage environment).

Who it’s for

Content creators, marketers, and users of coding agents who want to produce professional-looking, fast-paced educational or promotional videos without manual video editing software.

Highlights

  • End-to-end automation: Handles everything from script to final .mp4.
  • Multi-modal inputs: Supports starting from a text topic, an existing video, or a single image.
  • Human-in-the-loop: Includes decision gates for approving the narrative structure and visual style.
  • Integrated AI Stack: Leverages a variety of specialized models for image generation, video animation, voice cloning, and music generation via Atlas Cloud.

Sources