iOS & iPad · CarPlay · 100% On-Device

Any book. Any podcast. Any video.
Smart. Offline. Yours.

Text gets a voice — books, articles, web pages read aloud by 43 built-in neural voices in 21 languages, plus every Apple voice on your device — often 80+ in total. Audio gets a text — podcasts, audiobooks, lectures transcribed, punctuated and structured into paragraphs. Live translation in any direction. All on-device, no cloud, no account.

📱 Download Free See how it works
Books & EPUB Audiobooks Podcasts Web Video Voice Notes 🚗 CarPlay 🔒 No Account 🆓 Free to start 🍎 Apple NLP & ONNX
Scenes view
Karaoke mode Library Immersive word cycling Cross-library search
Library light theme

The medium changed. The line didn't.

Giving audio a transcript — or giving text a voice — doesn't change how you consume it. You still move through it in a straight line, losing hours just to reach the meaning.

SayFable reads differently. It weighs every word, so what matters rises and the filler falls away. It turns every sentence into a vector, so it can measure not just meaning but emotion. Together they break the text into scenes — self-contained blocks of meaning — a map you can navigate.

See what it's about in seconds. Jump to any point by what it means. Follow the emotional arc. Search by keyword and by sense. Gather the moments that matter into a playlist.

Find what matters to you — and share it.
Say your fable.

Drop it in.
Get back an intelligent object.

File or link — two minutes later: scenes, characters, voices, emotional arc, importance scores. A complete semantic map of the text, computed entirely on your device.

Import in progress — Preparing scenes

Importing…

Atmosphere — overview stats Atmosphere — scene bars

Analysed

Sayfable runs a three-phase analysis on every book — scene segmentation, entity recognition, emotional annotation. The map is ready the moment import completes — scenes, characters, scores, all waiting.

781
Paragraphs
370
Scenes
0.63
Emotionality
0.74
A/D Ratio
🎭
Voice Casting
Every character detected at import gets a name, gender, and voice slot. Assign each one a voice, hit "Save & Annotate" — and synthesis turns your book into a fully-cast radio play. No other TTS app exposes this.
🗺
Scene Structure
Chapters split into sequential scenes — each scored by importance, sentiment and structural type: action, dialogue, exposition or mixed. Find the most emotional moments, the most important passages, the most dynamic scenes — at a glance, before playback.
🎨
Generated Cover Art
Unique artwork per content type — vinyl disc for Music, layered pages for Text, broadcast arc for Speech, waveform ring for Record.
📦
.say Export
Your analysis, scenes, playlists, voice assignments and speaker corrections — everything you build, captured in one compact, portable file.
Composer

One object. Any sources.

Combine a web article, three podcast clips, a chapter pasted from a book, and two local audio files. Sayfable unifies them into a single seamless object — audio transcribed to text, text synthesised to audio, foreign language translated to yours. Play it from start to finish as if it was always one thing.

🎙 Podcast link
📄 Pasted excerpt
🎵 Local audio file
🌐 Web article URL
📻 Another episode
One object
Audio → transcript
Text → expressive voice
Foreign language → translated
Key scenes by importance
Most emotional paragraphs first
Level 1
Object
📖 Book 🎵 Music album 🎙 Audiobook
Level 2
Segment
Chapters Tracks Chapters
Level 3
Element
Paragraphs 15–25 s clips Paragraphs
Projects — group objects into a context. One tap to switch the whole library scope.
Scenes — span elements inside an object. No extra layer — just a named range stored on the object itself.

Every text in every object is editable — fix a transcription error, cut a repetitive intro, reorder chapters. Your object, your edit.

From EPUB to
intelligent audiobook.

Import, analyse, navigate, search, synthesise, translate — the complete Sayfable workflow on a real book. Jump to any chapter.

0:00 📥
ImportEPUB upload · chapter & speaker selection · language choice
0:42 🧠
AI AnalysisScene segmentation · character detection · voice casting
1:16 🗺
Scene NavigationBrowse scenes per chapter · play any scene directly
1:44 📊
Book AtmosphereImportance & emotion metrics · extracted characters · locations · speech parts
2:52 🎵
Smart PlaylistChapters by importance · scenes by type, emotion & importance · load & play
3:51 🔍
Semantic SearchFilter by character · keyword · select scenes → load playlist
4:36 🌐
Player · Word HighlightTTS synthesis in English · word-by-word highlight · live translation to Spanish
5:50 🇪🇸
Spanish PlaybackBook text & Spanish neural voice — full bilingual experience

Text that breathes.
80+ voices. All offline.

Every sentence synthesized with its own pitch, speed, and pause — shaped by the analysis. Dialogue softens and slows. Action peaks louder. Questions rise. All parameters baked into audio files before playback — no live synthesis, no dropouts, CarPlay-ready. Apple TTS and Silero neural voices across 20+ languages.

Expressive Reading ON
Ava Premium → Eo Jamie → Darrow dialogue 0.90× pauses shaped
Expressive Reading OFF
Jamie Premium only uniform rate no pauses

Same passage — Red Rising, the Darrow and Eo kitchen scene. With Expressive Reading ON, voices switch automatically by speaker. Pauses appear at emotional beats. No manual configuration. Works for every sentence, every chapter, every book.

Papa's voice. Any book. Any bedtime.

Record 15 minutes of your voice once. Sayfable uses Apple's Personal Voice to synthesise any text in that voice — books, articles, stories, notes. A dad can read every bedtime story his child will ever want, from any book, on any night. Even when he's not there.

Personal Voice · Red Rising — Darrow & Board of Quality Control Eduard's Voice
0:000:15

Synthesised on-device · Apple Personal Voice · no cloud, no upload

🌙
Screen dims. Story plays. Phone stays awake.
Apple's privacy rules mean Personal Voice is always synthesised live — it can't be pre-recorded. So when the phone sleeps, the story stops. The Stay Awake Timer solves this: set 30 minutes, the screen goes dark, the room stays dark, and the phone sleeps exactly when the story should end.
No known alternative

Also: phone in pocket, video playing, screen dark — stays awake for the whole commute.

Say hello in your language.

Every voice runs entirely on-device. Languages no other iOS app supports — including Latvian, Ukrainian, Belarusian, Kazakh, Uzbek, Tajik, Kyrgyz, Azerbaijani, Armenian, Georgian and more. For Ukrainian and Belarusian: neural stress prediction with 99% coverage so synthesis sounds natural, not robotic.

🇱🇻
Latvian
Sveiki! Laipni lūdzam SayFable!
🇺🇦
Ukrainian
Привіт! Ласкаво просимо до Sayfable.
🇧🇾
Belarusian
Прывітанне! Запрашаем у Sayfable.
🇰🇿
Kazakh
Сәлем! Sayfable-ге қош келдіңіз.
🇺🇿
Uzbek
Salom! Sayfable'ga xush kelibsiz.
🇦🇿
Azerbaijani
Salam! Sayfable-ə xoş gəlmisiniz.
🇦🇲
Armenian
Բարև! Բարի գալուստ Sayfable.
🇬🇪
Georgian
გამარჯობა! კეთილი იყოს Sayfable-ში.
🇰🇬
Kyrgyz
Салам! Sayfable-ге кош келиңиз.
🇹🇯
Tajik
Салом! Хуш омадед ба Sayfable.
🏔️
Bashkir
Сәлам! Sayfable-ға рәхим итегеҙ.
🏔️
Tatar
Сәлам! Sayfable-га рәхим итегез.
🏔️
Kabardian
Сэлам! Sayfable-м фыкъеблагъэ!
🌿
Sakha (Yakut)
Дорообо! Sayfable-га кэлбиккит.
🌿
Khakas
Сал! Sayfable-ке тохтапча.
🌿
Kalmyk
Менд! Sayfable-д тавтай морилтн.
🌿
Chuvash
Салам! Sayfable-па паллаштарăп.
🌿
Udmurt
Зечбуресь! Sayfable-э лыктэм.
🌿
Erzya
Шумбрат! Sayfable-с пракадовть.
🌿
Moksha
Шумбра! Sayfable-с кандовть.

Edit. Correct. Perfect.
Save everything.

The analysis is a starting point — you have full editorial control over the result. Fix a mistranscribed word, merge split characters, assign a specific voice to a single sentence. Every correction is preserved and exports with the book.

✏️
Editable Text
Every paragraph editable directly in the player. Remove front-matter from books. Fix STT inaccuracies in transcripts. 80–90% accuracy is enough for navigation — the rest is yours to correct.
🎭
Characters & Voice Casting
Sayfable detects every named character — gender, frequency, all aliases. Merge split entities ("Long John Silver", "Silver", "the pirate" → one character). Assign any Apple or Silero voice to any character. The narrator speaks in one voice; every character in another.
🎙
Per-sentence Voice
Assign any voice to any individual sentence. Automatic detection covers most dialogue — but you can go further and build a full radio play with every line exact.
🔤
Stress Dictionaries
For Russian, Ukrainian and Belarusian — add any word to your personal stress dictionary. Wildcard entries fix all declension forms at once. Your entries always override the built-in models.
Character Studio · Red Rising
.say

Your radio play. Your playlists.
Yours to share.

You wrote the casting — narrator in one voice, each character in another. You built the playlist of scenes that matter most. You added a parallel translation so a friend can hear Tolstoy in their language. .say saves all of it — voice assignments, custom stress corrections, playlists, scene names, translations — in a single file. No waiting for a publisher. No waiting for an audiobook studio. Open it on any iPhone and it reconstructs exactly as you built it, instantly.

Character voices & narrator assignment
Smart playlists — most important, by character, by emotion
Parallel translation for any language
Custom stress corrections
Renamed scenes & editorial notes
Share with friends — they get your version, ready to play

No internet, no server, no rights issues. The .say file contains only your data — the source stays on YouTube or your device. Recipients stream from the original source with your intelligence on top.

Any URL becomes
an intelligent object.

YouTube video, podcast, web article — paste the URL. Sayfable adds paragraph navigation, on-device translation, neural dubbing and TTS offline listening. News, lectures, tutorials: in your language, at your pace, with your personal voice.

▶️
YouTube stays in the app
Video streams from YouTube's own servers — no download, no ToS violation. Sayfable adds intelligence on top.
🔊
Live cast translation
In the demo: Spanish journalism, translated to English. The subtitle text appears in English as the video plays — tap any paragraph to jump. Enable dubbing and you hear the English synthesis mixed with the original Spanish voice, with independent volume controls for each track.
📖
Navigate like a book
Tap any paragraph to jump. Search by meaning across all videos — not just the exact timestamp with the exact word.
🎧
Listen offline — via synthesis
Switch to TTS mode: Sayfable reads subtitle paragraphs with your neural voice offline — news, lectures, tutorials, like a book.
🔗
.say = share without rights
Share your annotations as a .say file — URL only, no audio. Recipients get your navigation and translation; video streams from YouTube for them too.
🌐
Translation · Apple on-device · any content

Works across all content types — books, podcasts, recordings, YouTube. Three modes: text only (read the translation), audio only (hear the synthesis in your language), or both simultaneously. For Russian and other languages without Apple sentiment analysis, the English translation also improves scene importance scoring.

Subscribe once.
Stay current forever.

Paste a channel, podcast link, or news site — Sayfable keeps it fresh. New episodes, articles, and posts arrive automatically. No algorithm, no feed to scroll. Your library, quietly updating itself.

▶️
YouTube Channel
Subscribe to a channel
Paste any YouTube channel URL. Sayfable fetches the Atom feed, grabs the latest videos, and syncs subtitles automatically.
  • Auto-fetch new videos in a freshness window
  • Subtitles load with 1 s inter-episode delay
  • Check Now — delete stale, pull fresh
  • Scene building after all segments arrive
youtube.com/@channel → Atom feed via FeedKit
🌐
News & Articles
RSS that finds itself
Paste a bare homepage — meduza.io, dw.com, any news site. Sayfable auto-discovers the RSS feed from <link rel="alternate"> and imports it as a subscription. No wrestling with feed URLs.
  • Auto-discovery from bare homepage URLs
  • RSS description first — full article scraped lazily
  • Language auto-detected from scraped body
  • Perfect for news, blogs, serialised fiction
meduza.io → /rss → subscription
🎙
Podcast
Episodes that transcribe themselves
Subscribe to any podcast RSS feed. New episodes download, transcribe in the background, and become fully searchable — scenes, characters, importance scores. Audio streams from CDN after transcription.
  • Auto-download + background STT transcription
  • Captions-URL fast path (<podcast:transcript>)
  • Stream from CDN — local file deleted after transcribe
  • Keep audio files locally (optional)
podcast.rss → MP3 → transcribed speech object
▶️
One paste, perpetual sync
Paste a channel once. Sayfable checks for new content on launch and periodically in the background — no manual refresh needed.
🌐
Auto-discovery for news
Paste a bare homepage URL. Sayfable finds the RSS feed automatically — no feed URLs, no guesswork.
🎙
Podcasts become searchable
Background transcription turns every episode into a fully searchable, scene-structured object. Same semantic analysis as books.
🗞
Your daily news digest — spoken
Sayfable pulls the key stories from your feeds, summarises them, and reads the briefing aloud in a neural voice — hands-free, like a morning radio bulletin. It refreshes only when enough new articles arrive.

Any audio becomes
navigable text.

Podcast, audiobook, lecture, interview — transcribed on-device with neural punctuation. No 30-minute cap. No upload. After transcription, audio navigates exactly like books: scenes, semantic search, smart playlists.

English German Spanish Russian + 30 languages via Apple STT

Neural punctuation (Silero TE · offline): EN · DE · ES · RU

01
On-device STT Apple SpeechAnalyzer (iOS 26+) or SFSpeechRecognizer — word-level timestamps, gap recovery pass fills silent segments.
02
Neural punctuation Silero TE model (DistilBERT + LSTM, 85 MB, fully offline) restores commas, periods, question marks and capitalisation.
03
Semantic analysis Same pipeline as books: scenes, characters, importance scores, search index — built automatically.
04
Editable result Fix any inaccuracy directly in the player. Correct once — re-synthesis picks up the fix automatically.

Drop it in. From anywhere.

Sayfable accepts content from wherever you find it. A book file, a podcast link, a web video URL, an audiobook folder, a webpage, your own voice.

📚
E-Books
EPUB · FB2 · PDF · TXT
🎧
Audio Files
M4A · MP3 · WAV · FLAC
🎙
Podcasts
RSS · Direct link
▶️
Web Video
Any public URL
🌐
Web Articles
Any webpage
🎤
Your Voice
Record · live preview transcription

One slot, forever free.
Add more when you want.

All AI runs on your device — no servers, no tokens. We give you one slot free, forever. Buy slots one by one when you need them. Or get unlimited everything with Ultra — less than a single paperback per month.

FREE — forever

Your first slot: full AI, no limits

Import one book, podcast, YouTube video or article. Get everything: neural voices, emotional arc, character map, semantic search, scene navigation, transcription, translation, dubbing. No time limit. It stays yours forever — even if you never pay.

✓ Full media player — all five types
1 object: full AI + neural voices forever
✓ Semantic search, scenes, dubbing, translation
✗ Other objects: basic playback only
✗ No cross-library semantic search
✗ No .say export
Shelf · Pay once per object
Not ready for a subscription?
Buy one slot at a time.

Each slot is two things at once: the right to create a new object with full AI, and the right to preserve everything you build inside it — scenes, voices, playlists, translations — and take it with you. One-time, no recurring fee. The enrichment is yours forever. When you find yourself buying the third or fourth slot, Ultra will feel obvious.

€1.99
per object · one-time
3 slots = €5.97
Ultra = €5.99/month · unlimited
Ultra — unlimited everything
Ultra Yearly
39.99
per year · save 44%
  • ✓ All features from Monthly
  • ✓ Less than €3.34 / month
  • ✓ 2 months free vs monthly
  • ✓ Speechify charges $139/year for less
Discounts
🎓
Students & Schoolchildren €30 / year Annual only. Self-declared. No ID required. We trust you.
👴
Seniors 60+ €30 / year Annual only. Self-declared. No verification needed.
Available in Settings after onboarding.

Your data never leaves your device.

🔒 What we guarantee

Speech recognition, synthesis, translation, semantic analysis — all on-device. No cloud, no uploads, no model training on your data. Even your subscription status is a local unlock. Delete any book and every trace disappears instantly.

⚖️ Your responsibility

You own the content you import — personal books, purchased ebooks, public domain works. Sayfable is a reading tool, not a content host. Personal use only.

DMCA / Contact →

Operated from Latvia, EU · GDPR compliant · No third-party analytics
Silero TTS, Silero Punctuation and AccentorEngine models are used under the MIT License · © Silero Team

Start with any book.
See it in 2 minutes.

Import any EPUB, paste a podcast link, and you'll see the emotional arc of the whole book before you finish your coffee.

🆓 Free to start ⚡ 2 min full analysis 🔒 100% on-device 🍎 iOS 18+
📱

Sayfable is coming soon to the App Store.

Check back soon!