Transcription, Follow Along & Recap

Read every word as it's narrated with on-device transcription in the book's own language, follow the spoken line in real time, generate spoken recaps when you've zoned out, and skip non-narrative stretches like LitRPG stat blocks — all processed privately on your device.

Overview

FableFrog can turn the audio you're listening to into text, entirely on your device. This unlocks a family of features that share one foundation:

• Transcription — a faithful, word-by-word record of the narration, built in the background while you listen.
• Follow Along — a synced transcript reader that highlights the spoken line and scrolls itself as the narrator reads.
• Recap — a short written summary of the last few minutes, for when you zoned out and need to re-orient.
• Spoken Recap — that same recap read aloud, hands-free, so you can catch up while driving.
• Content Skipping — automatic detection of non-narrative stretches (like LitRPG stat blocks) that you can tap to skip past.

Everything here runs on-device. Audio is never uploaded, no transcript leaves your phone, and there are no servers or accounts involved beyond your own Audiobookshelf. This is consistent with FableFrog's "Data Not Collected" privacy posture — see Privacy & Security.

Requirements

These features have stricter requirements than the rest of FableFrog:

• iOS 26 or later. Transcription and content analysis rely on system frameworks that ship with iOS 26. On earlier iOS versions these sections simply don't appear; the rest of FableFrog works as normal.
• Premium. Transcription, Follow Along, Recap, and Content Skipping are part of FableFrog Premium.
• Apple Intelligence (Recap only). Spoken and written Recaps are generated by Apple's on-device language model, so Recap additionally requires a device that supports Apple Intelligence with it enabled. Transcription, Follow Along, and Content Skipping do not need Apple Intelligence.

How Transcription Works

When you open a book that has downloaded audio on an eligible device, FableFrog transcribes it in the background while you listen, using Apple's on-device speech recognition. The result is stored in a durable transcript bank — one file per book, kept alongside the downloaded audio.

A few things make this reliable rather than a gimmick:

• It works ahead of the playhead. The engine prioritizes transcribing the part of the book just ahead of where you're listening, so Follow Along and Recap have the words they need by the time you reach them.
• It survives sign-out. Because the bank mirrors your downloaded audio, your transcripts aren't wiped when you sign out — re-download isn't required to get them back.
• It only transcribes what it can read. Transcription works from downloaded audio. If a download is incomplete or corrupted, FableFrog tells you so instead of silently failing, and you can re-download to try again.

You'll see a Transcribing… badge with a live percentage while a book is being processed. Keep the book's detail page or the transcript reader open and transcription runs faster, because the device stays awake and focused on it.

Managing Transcription Per Book

Open a book's detail page and find the Transcription section (visible on iOS 26+ for Premium users). From here you can:

• See transcription progress as a percentage.
• View Transcript — open the full Follow Along reader.
• Export Transcript — save the transcript as Plain Text, Timestamped Text, Subtitles (SRT), or Subtitles (VTT) via the iOS share sheet.
• Redo Transcription — discard and rebuild the transcript from scratch.
• Remove Transcript — delete the banked transcript and opt this book out. The opt-out is durable: a removed book stays off until you explicitly tap Transcribe This Book again.

Power & Thermal Safeguards

Transcription is CPU work, so FableFrog is careful about when it runs:

• On battery, only the book you're currently playing is transcribed. To transcribe a book you're not playing, plug in, play that book, or set Transcribe to Always in Settings.
• It pauses when your device gets warm and resumes automatically once it cools down.
• It pauses in Low Power Mode and resumes when you turn Low Power Mode off.

When transcription is idle, the Transcription section explains exactly why (for example, "Paused — Low Power Mode is on"), so a stuck percentage is never a mystery.

You control the overall policy in Settings → Playback, under the transcription power setting — choose between transcribing only while plugged in / playing, or Always.

Transcription Language

FableFrog transcribes each book in its own language. A German audiobook is transcribed by the German speech recognizer, not the English one.

The book's detail page shows which language is in use, right under the Transcription section - for example "Transcribed in German".

How the Language Is Chosen

FableFrog picks the language most-specific-first:

1. Your per-book override, if you've set one. This always wins.
2. The language in your Audiobookshelf metadata for that book. FableFrog reads this field generously, so de, de-DE, ger, German, and Deutsch all resolve to German.
3. English, only when the metadata is missing or unrecognizable.

That last point matters: English is the fallback for a language FableFrog couldn't identify, never a substitute for one it could. If your book is in a language Apple's speech recognition doesn't support on your device, FableFrog tells you so and skips the book rather than transcribing it in English and labeling the result as something it isn't. A transcript you can't trust is worse than no transcript.

Overriding the Language

If your server's metadata is wrong or blank, you can correct it for a single book. On the book's detail page, tap the language row in the Transcription section and pick a language from the menu, or choose Automatic to go back to metadata-driven behavior.

Changing the override re-transcribes the book in the new language automatically. There's no separate "redo" step to remember.

Language Models

Apple's speech recognition needs a model for the language it's transcribing.

• Already installed - transcription starts right away.
• Available but not installed - FableFrog downloads it for you in the background and picks the book up once it's ready. You'll see "Downloading German language model…" while that happens. These downloads respect your Wi-Fi Only setting, the same control that governs book downloads.
• Not supported on your device - the book's detail page shows "German unavailable on this device" and the book isn't transcribed.

Books Transcribed Before This

Books that were transcribed under the old English-only behavior are detected and healed. A non-English book holding an English transcript is re-transcribed in its correct language, and if that language isn't available on your device the misleading transcript is discarded rather than kept.

Books that were genuinely English are left completely alone - there's no mass re-transcription to sit through.

Follow Along

Follow Along is the synced transcript reader. Open it from the View Transcript button on a book's detail page, or from the player's menu while listening.

• The currently spoken line is highlighted as the narrator reads it.
• The reader auto-scrolls to keep the current line centered.
• When you scroll up to read ahead or look back, auto-scroll pauses and a Resume following pill appears. Tap it to snap back to the playhead and start following again.
• Opening the reader jumps straight to the current position in the book, so you're never hunting for your place.

The highlight and the playhead stay in lockstep, which makes Follow Along useful for dense passages, character-heavy scenes, unfamiliar names, or listening in a second language.

Recap

Zoned out for a few minutes? A Recap is a short written summary of what was just narrated, generated on-device to get you re-oriented without rewinding and re-listening.

• Open it from the Recap control in the player.
• Choose a window of the last 5, 10, or 15 minutes of book time, ending at your current position.
• FableFrog summarizes the transcribed words in that window using Apple's on-device language model.

A Recap is distinct from the verbatim transcript: the transcript is every word, the Recap is the gist. Recaps are generated fresh each time and are not stored.

A recap comes back in the language the book was transcribed in - a German book gets a German recap - and the recap labels its language so you know what you're reading. See Transcription Language.

The Shape of a Recap

A recap is two parts, not a wall of prose:

• Where things stand - one or two sentences setting the stage as the passage opens: who these people are and what larger thing they're in the middle of.
• What happened - an ordered, bulleted timeline carrying you from there to your current position. The list is longer for a longer window, and it favors the things that move a story - decisions, arrivals, discoveries, injuries, things going wrong - over scenery and repeated song lyrics.

The two parts deliberately describe different moments, which is what stops the second half from restating the first. FableFrog renders both section headers itself rather than asking the model for them, so the structure holds in every language.

Earlier Context

The opening line used to see only what was inside your recap window. Ask for a 10-minute recap when the characters set out on a journey 20 minutes ago and it couldn't mention the journey at all.

FableFrog now reads up to 20 minutes of listening before your window, condenses it on-device into a short background note, and grounds the recap on it. You'll see a brief Setting the scene… state while that runs on a cold open.

Two guarantees hold:

• No spoilers. Nothing past your current position is ever fed to the model, for the background note or the recap. A recap cannot reveal something you haven't heard yet.
• Never worse than before. If the background pass fails or has nothing to work with - near the start of a book, for example - you get the window-only recap that worked before, not an error.

Those condensed notes are cached per book, so re-opening the sheet or nudging the playhead doesn't pay the cost a second time. They're cleared alongside your transcripts on sign-out.

Spoken Recap through Siri uses a cached note if one is already warm and otherwise stays window-only. It never runs the condensing pass itself, because a Siri command that stalls for fifteen seconds before speaking isn't worth the extra grounding.

When Recap Isn't Available

The Recap control tells you why it can't run, rather than just greying out:

• Apple Intelligence is off — turn it on in iOS Settings to enable Recap.
• The model is downloading — Apple Intelligence is still preparing; try again shortly.
• Not enough transcript yet — if the last several minutes haven't been transcribed (for example, you just opened a brand-new book), there are no words to summarize. Give transcription a moment to catch up.

If your device doesn't support Apple Intelligence at all, the Recap control simply doesn't appear.

FableFrog re-checks all of this during a session rather than only at launch. Start the app while Apple Intelligence is still downloading its model and Recap turns itself on the moment the model lands - no relaunch. Turn Apple Intelligence off mid-session and Recap notices that too.

When a Recap Fails

Three things can stop a generated recap, and they need different responses, so FableFrog tells them apart:

• The passage was too long. A fast-narrated 15-minute stretch can overflow the on-device model's context window. Size is the problem, so FableFrog retries automatically with a shorter, most-recent slice.
• The content was refused. Apple's on-device guardrail screens the passage going in as well as the summary coming out, and ordinary genre-fiction violence used to trip it - a LitRPG or grimdark title could fail every single time. Recap now runs under the relaxed guardrails Apple ships for transforming text you already own, which fixes the common case. A refusal is still possible, and when it happens the sheet says so and points you at the window picker. It deliberately offers no Try Again: every smaller slice of the same passage would be refused identically.
• The model was unavailable. Sometimes the on-device model stops answering, and the problem is the model rather than your book. That case now says so instead of hiding behind a generic failure, and its Try Again is worth tapping - FableFrog rebuilds its connection to the model first, so the same passage can succeed the second time. It used to be that one such failure took every recap after it down with it until you force-quit the app. It doesn't any more.

Spoken Recap

Spoken Recap is the same recap, read aloud on-device — built for when you can't look at the screen, like driving.

• Trigger it from the Recap screen, or hands-free with Siri: FableFrog ships a Spoken Recap App Intent, so you can ask Siri for a recap without touching your phone.
• While the Spoken Recap plays, it takes over audio output: if the book was playing, it pauses, the recap speaks, and then the book resumes automatically. If the book was already paused, it stays paused afterward.
• A resume triggered this way doesn't apply your usual rewind-on-resume, since the recap has already caught you up.

The spoken text is identical to the written Recap — it's the same summary, just narrated.

Content Skipping

Content Skipping finds stretches of a book you may want to jump past — classified by what kind of content they are — and lets you skip them with a tap. It reads the banked transcript text (never the audio again), so it's fast and private.

In this release, the one shipping category is LitRPG stat blocks — the recurring tables of stats, levels, and inventory that punctuate LitRPG and progression-fantasy audiobooks. In the app this category is surfaced as Skip Stats.

How Skipping Works

• As your playhead nears a detected stretch, the player shows a skip button on the cover art.
• You tap to skip — FableFrog never skips automatically. Every skip is your call, so a false positive just means an ignored button, never lost narration.
• Skip regions are recomputed as more of the book is transcribed, so coverage improves while you listen.

Choosing Which Books Are Analyzed

Content Skipping resolves whether to run on a given book from three controls, most specific first:

1. Per-book override — turn analysis on or off for a single book. This always wins.
2. Author allowlist — force analysis on for specific authors (useful for LitRPG authors whose every book has stat blocks).
3. Global toggle — the default applied to all books when neither of the above is set.

All three sit beneath the Premium and iOS 26 requirements. The global toggle is the default for all books, not a master kill switch — a per-book or per-author setting can still override it.

Exporting Transcripts

From a book's Transcription section, Export Transcript writes the full transcript to a file you can share or save:

Exports are generated on-device and shared through the standard iOS share sheet — nothing is uploaded.

Privacy

Every part of this feature set is local:

• Audio is transcribed on your device using Apple's on-device speech recognition; nothing is sent to a server.
• Transcripts are stored on your device, alongside your downloaded audio.
• Recaps are generated by Apple's on-device language model and are not stored or transmitted.
• There are no FableFrog accounts, analytics, or telemetry involved.

See Privacy & Security for FableFrog's full privacy stance.