Does Your TTS App Really Work in Airplane Mode?

You are at 30,000 feet with a PDF you meant to read last week. You open the text-to-speech app you installed precisely for this moment, paste in a paragraph, and press play. Nothing happens — or a spinner turns forever, or a banner tells you to check your connection.

The app’s App Store page said “offline.” It was not lying, exactly. It just meant something different from what you assumed.

Short answer: “Offline” is used for three incompatible things in this category. Some apps generate brand-new speech on the device with no network at all. Some only play back audio you converted earlier while you were online. Some have no offline mode whatsoever. Before a flight, the question worth asking is not “does it work offline” but “can it read text I hand it while I am already disconnected?”

Scope and sources: This article covers nine iPhone apps: Local TTS - Voice Aloud Reader, SpeakCove, PageEcho, VoiceReader AI, Speech Central, Voice Dream Reader, ElevenReader, Speechify, and NaturalReader, plus Apple’s built-in Spoken Content. Local TTS is our own app, so every claim below is taken from each product’s own published description — store listing, pricing page, or help documentation — rather than from our opinion of it, and each is quoted so you can check it against the source. Listings were checked on August 23, 2026. Apps change their plans often — confirm the current listing before paying.

The three things “offline” means

Sort the category this way and the confusion disappears.

1. Live on-device synthesis. The speech model runs on your iPhone. You can paste text you have never seen before, while disconnected, and hear it read. This is the only meaning that survives a flight.

2. Pre-generated audio playback. You convert text to an audio file while connected, the file is stored on the device, and you play it later without a connection. This is genuinely useful — it is how podcasts work — but it is offline listening, not offline speech. New text gets you nothing.

3. No offline mode. Every request goes to a server. Without a connection, the app is a document viewer.

There is also a second axis that catches people out, and it applies even to category 1: what else needs the network. A voice model usually has to be downloaded once. A subscription has to be validated. A cloud library has to sync. An app can synthesize perfectly well on-device and still be useless on a plane because you installed it at the gate and never downloaded a voice.

What each app says about itself

Everything in this table is quoted or paraphrased from the product’s own store listing, pricing page, or documentation.

Read down the first column and the category sorts itself: seven of these ten generate speech locally, one generates it remotely and caches the result, and two need the network for every request.

The ElevenReader case, because it is the clearest

ElevenReader is worth pulling out because it is the most common source of the misunderstanding, and because its own documentation is admirably precise about it.

ElevenReader’s pricing page describes offline downloads as a feature that lets you “convert text into audio ahead of time so you can listen without an internet connection.” Every constraint follows from that one sentence:

• The conversion needs the network. Only the playback is offline.
• It is an Ultra plan feature. The free tier, which includes 10 hours of text-to-audio per month, generates through the network.
• Downloads are capped at roughly 20 hours per day.
• Files are stored for up to 60 days and are removed when you sign out.

None of that is a flaw. ElevenLabs runs large models that do not fit on a phone, and pre-generation is the honest engineering answer to that constraint. It is simply a different product from an app that synthesizes locally, and the word “offline” hides the difference.

The one that surprises people: Voice Dream Reader

Voice Dream Reader’s listing says “all voices work offline,” and its scanner “works entirely on device: no need for internet.” Both are true, and it has been the accessibility community’s default for years.

But the same listing also says “a subscription is required to access the app.” Speech generation being local does not help if the app gates itself behind a subscription. This is the second axis from earlier: a category-1 engine with a category-3 dependency bolted to the front door. The two claims are not in conflict — the voices really are local — but “works offline” and “opens offline” are separate questions, and the listing only answers the first.

What still needs a connection, even in the best case

If you are packing for a flight, these are the four things that catch people regardless of which on-device app they chose:

1. First-run voice downloads. VoiceReader AI states this plainly: offline works “after the initial voice download.” Most on-device apps ship the app small and fetch the model separately. Install and open every app before you leave.
2. The specific voice you want. Downloading one voice does not download all of them. Switching to a different narrator mid-flight can fail even when playback works.
3. Subscription validation. Any app requiring an active subscription may re-check it.
4. Cloud document libraries. An app that stores your files on a server has your files on a server. Import to local storage first.

The thirty-second check, before you rely on any of them

Documentation tells you what a product intends. Your own phone tells you what it does on the day you need it. The check is short:

1. While still connected, open the app, finish first-run setup, and download the voice you plan to use.
2. Turn on Airplane Mode — then confirm Wi-Fi is also off. Airplane Mode leaves Wi-Fi on if it was already on, which quietly invalidates the whole exercise.
3. Paste in a sentence the app has never seen and press play.

Step 3 is the entire question. Playing a document you imported earlier proves only that a file was stored. Speech from text the app is seeing for the first time, with no network, is what separates category 1 from category 2 — and it is the thing you will actually be doing at 30,000 feet.

If it fails, read the error carefully. “Network connection required to generate speech” tells you the engine is remote. “Voice not downloaded” tells you the engine is local and you skipped step 1.

Limits of this comparison

A few honest boundaries, because a comparison that claims to settle everything settles nothing:

• This is a documentation comparison. Every row reflects what a product publishes about itself as of August 23, 2026. Published claims are a reasonable starting point — they are what a vendor is accountable for — but they describe intent, not your specific device on a specific day.
• Offline is not the same as good. Large server models genuinely produce better audio than anything that fits in a phone. Choosing local means accepting that trade, and it is a real one.
• Plans change frequently. ElevenReader’s offline tier, Speechify’s word limits, and every price mentioned here can move next quarter.
• DRM-protected books are out of scope. No option in this list can read a Kindle or Apple Books purchase aloud, offline or otherwise. You need a DRM-free file you are authorized to open.

What we build, and why it is in this list

Local TTS is our app, so treat this section as disclosure rather than a verdict.

It sits in category 1. Speech is generated on the iPhone by Supertonic 3 for 31 languages and Kokoro-82M for its English voices, which means new text pasted in at altitude is synthesized sentence by sentence rather than fetched. There is no account, so there is nothing to log in to. On-device OCR means a photographed page also works with no connection. The free plan reads up to 500 characters per note — enough to run the thirty-second check above without paying anything, which is more convincing than this paragraph.

A reviewer on the US App Store put the category distinction more plainly than we have managed to:

“Compared to the pay walls like other TTS apps, being able to use it offline and simply copy and paste without limits is what other apps are missing.”

— #night11, five stars, August 21, 2026, App Store

The honest limits: it is iPhone and iPad only, it reads PDF, EPUB, DOCX, TXT, and Markdown rather than every format in existence, and its voices will not beat a large cloud model on pure naturalness. If your reading happens at a desk on Wi-Fi and you want the best-sounding narration available, a server-based app is a reasonable choice. If your reading happens on planes, trains, and subway platforms — or the documents are ones you would rather not upload — the local category is the one to shop in, and there are several good apps in it.

The one-line version

Before you fly: open the app, paste in a sentence you have never used, turn on Airplane Mode, and press play. Thirty seconds now beats an hour of silence later.