Universal ingest
EPUB, PDF, DOCX, RTF, TXT, Markdown, HTML, a URL, pasted text, even OCR for scanned images. If it's a book, it goes in.
Local audiobook studio
Your books,
performed.
Drop in any book and get back a high-quality audiobook, read by a narrator of your choice. Then shape every line in a real editor until it sounds right, all on your own machine.
Runs 100% on your machine. No account, no upload, works offline.
What it does
A whole audiobook studio, one drag away
From raw file to finished listen: ingest, narrate, edit, and export without leaving your desk or your privacy behind.
Single-narrator render
A clean, natural narrator with stream-as-it-renders playback and word-level karaoke highlighting so you can read along.
Narration Editor
The heart of the app. Click any word to fix a pronunciation, audition the line, or voice its sound effects, and see every change previewed before you render.
Voice cloning
Consent-gated cloning with optional per-emotion reference clips, so a narrator can carry the right tone across the whole book.
Runs on any GPU
NVIDIA (CUDA) or AMD (ROCm) on Windows, or plain CPU. Intel Arc (XPU) is experimental. The right runtime is picked for your machine automatically.
Full cast Preview
An experimental multi-cast mode is in progress: it tries to give each character its own voice. Speaker detection is rough today and needs hand-correcting, so single narrator is the way to go for now.
How it works
Three steps to a finished listen
1
Drop in a book
Import any supported file, paste a URL, or point it at scanned pages. NarrationOS parses chapters, headings and dialogue for you.
2
Narrate & refine
Pick a narrator voice, then open the editor to fix pronunciations, audition lines, and voice sound effects until it sounds right.
3
Render & listen
Playback starts as it renders, with karaoke highlighting. When it's ready, export tagged audio files to keep.
The headline feature
The Narration Editor
A production isn't a single button. It's the hundred small fixes that make a narration sound right, and the editor is where you make them.
- Click any word or line to edit it: reassign, respell, or fix it in place, right in the reader.
- Set a pronunciation and audition the line: hear the real synth before you commit to a full render.
- Detect SFX & stylized laughs so they're voiced as real sounds, not spelled-out letters. Your reader text is never changed.
- Manage names and book-wide pronunciations, search the text, and see every change previewed inline before you render.
Inside the app
Built for readers, not spreadsheets
A calm, dark, distraction-free workspace: library, editor and player in one place.
Real screenshots from NarrationOS.
Private by design
Your library never leaves your machine
NVIDIARTX / GTX
CUDAAMD RadeonRDNA3 / RDNA4
ROCmIntel ArcExperimental
XPUNo GPU?Always works
CPULocal-first, not cloud-locked
- No account, no sign-in, no subscription server.
- The narration model and the attribution LLM both run on your PC.
- Works fully offline once first-run setup is done.
How it compares
A studio you own, not a meter you feed
Cloud narration services charge by the character and keep your text on their servers. NarrationOS is a one-time purchase that runs on your hardware.
| NarrationOS | Other services | |
|---|---|---|
| Pricing | One-time purchase | Monthly / per-character |
| Runs locally & private | Yes | Uploads your text |
| Works offline | Yes | No |
| Line-level editing & audition | Built-in editor | Re-generate & hope |
| Full-cast, per-character voices | In preview | Rare / manual |
| Whole-book length | No cap | Quota-limited |
| Runs locally | Yes | N/A (their servers) |
| Use your output commercially | Allowed* | Tier-dependent |
*Subject to the third-party model terms noted in the license. See the EULA for details.
Questions
Good to know
What file formats can I turn into audiobooks?
EPUB, PDF, DOCX, RTF, TXT, Markdown and HTML files, a web URL, or text you paste in. It can even OCR scanned images of pages.
Is the narration a real human voice?
No. The voices are synthetic, generated by AI text-to-speech models running on your own machine, not recordings of a person reading. You can also clone a voice from a sample you have the rights to use. See our disclaimers for details.
Do I need a powerful GPU?
No. It still runs on plain CPU, just slower than real time. A GPU is only for faster renders, and NarrationOS supports NVIDIA (CUDA) and AMD (ROCm) on Windows. Intel Arc (XPU) support is experimental and not yet verified on Intel hardware.
Is it really private and offline?
Yes. The narration model and the speaker-attribution LLM both run locally on your machine, and once first-run setup is finished the app works fully offline. Your books and voices never leave your PC.
What happens on first launch?
The app sets up its AI runtime from the files it ships with. On some hardware (for example, non-NVIDIA GPUs) it downloads the matching components on that first run, which can transfer several gigabytes once. After that, launches are fast and offline.
Who owns the audiobooks I make?
You do. You keep full rights to everything you render, including for commercial use, as long as the source text and any cloned voice are your own or you have permission to use them. NarrationOS makes no claim on your output and takes no responsibility for how you use the app. Making sure you have the rights to the content and voices you feed it is up to you. More in our disclaimers.
How much is it, and where do I buy it?
Which platforms does it run on?
NarrationOS is a Windows desktop app. Check the store page for current system requirements.
Is NarrationOS a text-to-speech (TTS) app?
Yes. At its core NarrationOS is offline text-to-speech software: it reads your text aloud with an AI voice. The difference from a plain TTS reader is what surrounds it, a full audiobook pipeline that understands chapters and dialogue, plus an editor where you correct how individual words are pronounced instead of accepting whatever the model says first.
How does AI voice cloning work, and do I need consent?
You record or import a short voice sample and NarrationOS builds a cloned AI voice from it, which then narrates any book you like. Cloning is consent-gated by design: you confirm you have the right to use that voice before the clone is created. Cloning a real person's voice without their permission can be unlawful, so only clone your own voice or one you have clear permission to use. See our disclaimers.
Can I convert an EPUB or PDF to an audiobook?
Yes. Drop in an EPUB, PDF, DOCX, RTF, TXT, Markdown or HTML file and NarrationOS extracts the text, detects chapters and headings, and renders the whole thing to speech. Scanned PDFs can be run through OCR first, so image-only pages still become audiobooks.
How natural do the AI voices sound?
The narration uses modern neural TTS, so it sounds far closer to a human reader than old robotic speech synthesis, with natural pauses and sentence rhythm. It is still synthetic and not a substitute for a professional human narrator. Where a line lands wrong you can re-roll it, adjust expressiveness, or fix the pronunciation directly in the editor.
Does the text-to-speech work offline, without a subscription?
Yes to both. The TTS and voice-cloning models run on your own hardware, so after first-run setup there is no internet requirement, no per-character quota and no monthly fee. It is a one-time purchase, unlike cloud TTS services that bill by the word or by the hour of audio.
Can I fix words the AI voice mispronounces?
Yes, and this is the main reason the editor exists. Names, invented words and acronyms are where TTS usually falls apart. You can set a pronunciation for any word, hear it previewed in a real sentence before committing, and have it applied consistently everywhere that word appears in the book.
Can I use multiple AI voices for different characters?
A single narrator is the recommended path today and is what the app is tuned for. There is also an experimental full-cast mode that detects who is speaking and assigns different voices to characters. Speaker detection is still rough and needs hand-correcting, so treat it as a preview rather than a finished feature.
Can I sell the audiobooks I generate?
Yes, provided the rights are yours. You keep full commercial rights to what you render, so long as you own or are licensed to use the source text and any cloned voice. Public-domain books and your own writing are the safest ground. Note that some audiobook storefronts have their own rules about AI-narrated titles, so check the platform you plan to publish on.
Made by one person who loves books
NarrationOS is built and maintained by a solo indie developer. Clear bug reports get fixed fastest, features land as time allows, and the community shapes what comes next. If that's the kind of project you like to support, you'll fit right in.
Join the Discord
Give your books a voice
Get NarrationOS on Steam or itch.io and start narrating tonight.