ImmersivePage

Experience multi-voice reading

A 90-second demo — no file needed.

Browse the library or import your story

EPUB, PDF, Word, or manuscript — read with character voices or a single narrator. Or get started right now with existing titles.

Filter titles

Bring in your content

Choose where the text comes from, then how you want it read. Fiction and ebooks with dialogue use character voices; articles and essays use a single narrator.

New here? Hear a 90-second multi-voice sample — no file needed. Try immersion presets on Listen.

First, create a new document

Start a fresh document, then add a file or paste text below.

2. How should it be read?

Article or essay One narrator — news, reports, nonfiction, quick reads. Fiction Full cast — novels, ebooks, and dialogue-heavy stories.

Illustrations (optional) Chapter art and prompt drafts for Cinema mode

Generate chapter illustrations AI creates one image per chapter for Cinema mode. Use Google Gemini (free tier, no install), a local engine (Ollama FLUX, Automatic1111, or ComfyUI), or OpenAI GPT Image. Requires Pro. Generate image prompts Draft Copilot-ready prompt text for each chapter — no images are rendered. Uses Settings → Illustration LLM only (your Ollama text model). Google Gemini in Settings is used only when you also check Generate chapter illustrations above.

Book in a series?

For fiction only: is this a new book or a sequel with saved casting?

New book Detect cast from scratch (book 1, or a standalone novel). Sequel Reuse saved casting from a previous book, then process this file.

Analysis method

Smart cast detection finds characters and attributes dialogue automatically. Fix uncertain lines on the Cast or Listen step — no API key required.

How is dialogue laid out in this book? One speaker per paragraph Each paragraph (block ending with a line break) is one character's turn. Multiple quotes in the same paragraph belong to the same speaker — typical of many epubs and self-published files. Multiple speakers per paragraph Characters may alternate within a single paragraph — typical of traditionally typeset fiction where each new speaker starts a new paragraph.

Smart detection Recommended Fast and private. Uses quote patterns, speech tags, scene continuity, and voice fingerprinting — then surfaces uncertain lines for you to review.

Most books sound right after smart detection plus a quick pass in the review queue or in-scene editing on Listen.

Experimental AI overlay (optional)

Adds an LLM pass on top of smart detection. Local models (Ollama) often perform worse than heuristics alone; cloud models may help on some books but are not guaranteed to improve results. You pay your AI provider directly. Use the review queue and Listen editing to fix mistakes either way.

Enable experimental AI Runs an optional LLM pass after smart detection. Uses Ollama on this PC or a cloud API key you provide.

Provider API key For Ollama on this PC, leave blank or type ollama (not an OpenAI key). Model For Gemini use gemini-2.0-flash or gemini-2.5-flash (not gemini-flash-latest — that id is auto-mapped).

How much should the AI weigh in? Uncertain lines only Default for AI Smart detection runs first; the AI only sees chunks it was unsure about. Faster and usually safer than a full-book AI pass. Calculating… Full-book AI pass Sends every chunk to the AI and runs an extra validator on disagreements. Slowest and most expensive; try only if uncertain-lines mode is not enough. Calculating…

Times and costs are rough estimates for a typical novel (≈100 000 words). Smart detection always runs first and is free. AI charges depend on your provider and book length.

What's the difference, technically?

Both options run smart detection plus continuity and voice-fingerprint passes first. Uncertain lines only sends just the chunks smart detection was unsure about to the AI. Full-book AI pass sends every chunk to the AI and runs an extra validator that double-checks confident heuristic picks where the AI independently disagreed.

Advanced

Process only first chunks Each chunk is roughly one scene (~4 000 characters, about 2–3 manuscript pages). A typical novel is 100–180 chunks. Useful for previewing results on a sample before committing to a full-book run.

Reading voices

Automatic is the easy default — it picks the best voice you have ready. Expand a section below only if you want a specific provider.

Processing your book

Starting…

Initial analysis

Cast review

Assign voices to characters. Preview samples, then save before listening.

Select a character.

Book settings

Director preset

Narrator

Human-verified sample

After you fix every wrong dialogue line on Cast and Listen, mark this book as a gold sample. We compare Process-time speakers to your corrections for real-world accuracy stats.

No offline books yet

Chapter break

The End

Book appearance

Cinema backdrop

Highlighting

Text & display

Processing

Monthly usage

Bring in your content

Book details

Book in a series?

Analysis method

Reading voices

Processing your book

Story bible

Chapter image prompts

Cast review

Recommended voices (★ = best match · distinct preview only)

Book settings

Your book

Find in story