Concepts

Jump to bottom

Steven Dunn edited this page Mar 30, 2026 · 4 revisions

Concepts

Understanding how Audiobook Studio organizes data is key to a smooth workflow.

🏗️ Hierarchy

Library: The collection of all your projects.
Project: Represents a single audiobook or collection. Contains metadata (Author, Series, Cover).
Chapter: A logical division of a project. Contains the text and its generated audio.
Segment: A single sentence or paragraph within a chapter. This is the smallest text unit stored in the database.
Chunk Group: The displayed Performance/Production unit. One chunk can contain several adjacent stored segments when they belong together.
Character: A persona assigned to segments. Chapters are narrated by a "Narrator" by default, but you can assign specific "Characters" to dialogue.

🎙️ AI Voice Lab

Voice: A higher-level identity (e.g., "Dracula"). This is what you assign to Characters in your projects.
Variant: A specific stylistic or emotional performance of a Voice (e.g., "Main - Calm", "Main - Shouting").
Sample: High-quality .wav reference audio used to clone a Voice.
Engine: The synthesis path attached to a profile, currently XTTS (Local) or Voxtral (Cloud).

🔄 Generation Workflow

Analysis: The system scans your text for long sentences (over 500 characters) and automatically splits them to ensure high-quality TTS.
Queuing: When you click "Generate", displayed chunk groups are added to a background queue.
Synthesis: XTTS, Voxtral, or a mixed chunk-aware path processes each queued chunk based on its assigned voice profiles.
Baking: After all chunks in a chapter are generated, they are stitched into a master WAV and simultaneously encoded into a high-quality M4A chapter cache.
Assembly: Finally, the system performs Lossless Concatenation of the M4A chapter files into a standard .m4b container. Because the encodes are cached, rebuilding the audiobook is nearly instantaneous.

Home | Library and Projects | Voices and Voice Profiles