Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.loraverse.io/llms.txt

Use this file to discover all available pages before exploring further.

A Loraverse contains three layers.

Media

The actual generated assets. Frames as PNGs, takes as MP4s, reference images included.

Structure

The hierarchy rebuilt on disk. Scenes contain beats. Beats contain shots. Characters contain DNA and appearances.

Memory

The context that proves how each asset was made. Prompts, model settings, lineage, source references, dates.
A few specifics that travel with every Pack:
  • A README at the root, written for a human. Opens with a project-at-a-glance summary. Generations, days of production, most-used model, motion seconds, most-iterated shot, most-appearing character.
  • A manifest.json at the root, written for a machine. The complete structure of the Pack in a single file.
  • Per-folder sidecar JSON like scene.json, beat.json, shot.json, character.json, dna.json, appearances.json. Each holds the slice of the manifest relevant to that folder.
  • Self-describing filenames like nina-in-the-park_beat-01_shot-01_frame-01_hero_gemini-3-pro-image-preview.png. The filename tells you what the file is without opening it.
When an editor opens a Pack and looks at a file, the filename and sidecar already tell them what it is. No what was this? email back to you. When a legal or compliance request asks how an asset was made, the answer is in the Pack: model, references, prompt, lineage, timestamp. When a training pipeline reads the Pack, it can group by character, by scene, by model, without anyone manually preparing the data. The Pack is structured so the next person, or the next tool, doesn’t have to ask.