MakeAudiobookAudiobook Workbench

Source text to narrated audiobook, with cost checkpoints.

The app now persists each source as a server artifact, reuses prior cleanup work when inputs match, assembles narration-sized segments, and renders either a tiny audio sample or a background-packaged M4B audiobook with selectable OpenAI voices and a podcast RSS feed.

Live parsers

4

Paste, TXT, PDF, EPUB

Pipeline stage

05

Extraction through audio render

Narrator voices

13

Built-in OpenAI voices with manual previews

Reuse

Cached

Cleanup, assembly, and matching render jobs are reused

Ingest

Upload a book file or paste raw text

TXTPDFEPUB

Narrator voice studio

Choose a built-in OpenAI voice for audiobook narration. This selects a narrator voice; it does not clone an author or private voice.

OpenAI.fm samples

Selected narrator

Marin

Preview generation is on-demand so the app does not spend API credit automatically.

Extract first, then run the GPT cleanup pass. Once cleanup is saved, narration assembly and audio rendering reuse that artifact instead of uploading the source again.
Cost guardrails: extraction is local and free. Paid AI cleanup and audio render steps require confirmation against the current text-based estimate.

Build notes

What this build does right now

MVP+

TXT

Plain text and markdown sources

PDF

Digital PDFs with embedded text

EPUB

Reflowable ebooks with chapter HTML

Extracted text is stored as a reusable artifact. Cleanup results are cached by normalized source and model, narration assembly is local and reusable, and full audiobook packaging now runs as a background job with a merged `m4b` output and podcast RSS feed.

Pipeline output

Run extraction to start the pipeline

Run extraction to inspect normalized text and parser warnings. Once the source looks right, cleanup, assemble, and render a tiny audio sample.