AI Video Episode Generator

This is a command line tool that takes a single topic and produces a finished short form video episode with very little manual work. It writes a structured narration script, picks a visual direction that suits the content, breaks the script into production ready scene plans and then generates, narrates and renders the whole thing into a final MP4.

The hard part of automated video is consistency, so the pipeline keeps a character and environment bible that carries recurring characters and settings across scenes. A style detector chooses the visual direction from the content itself, and the choice can be overridden manually when a specific look is needed. Each scene is generated as its own video segment before composition.

Audio and timing are handled automatically. Narration is generated as natural voiceover, then subtitles are aligned to the actual speech rather than guessed from the script, so captions land on the right words. The final render adds transitions, burned in captions, background music and ducking so the music sits under the voice instead of fighting it.

The workflow is built for repeatable episode production. Multiple content presets cover different storytelling styles, voices, prompts and rendering behaviour are configurable, previously generated assets are reused when nothing has changed and a full regeneration is available when they have. A fallback rendering mode keeps the pipeline producing a usable episode even when generated footage is unavailable.