Listening (text to speech)

Athena can read what you are looking at out loud. It uses your phone's built-in speech engine by default — no model downloads, no network, no cost. The alternatives below only matter if that engine does not work on your device.

Getting started

1. Open something, bring up the toolbar and tap "Listen".

2. Adjust speed, pick a voice, and set a sleep timer if you want playback to stop on its own.

3. In EPUBs the sentence being read is highlighted and the page scrolls along with it. PDFs are read aloud without any on-page highlight.

4. Playback continues with the screen off and while Athena is in the background. The lock screen and notification shade show the cover and chapter, with controls to pause and skip a sentence back or forward.

If it does not work: run the speech self-check

Some Android phones — especially models whose manufacturer modified the system speech layer — have a broken speech engine. Go to Settings → Listening → Speech self-check. Athena will actually play a test phrase and ask you whether you heard it: the programmatic checks are unreliable on exactly these devices, so your ears are the final judge.

The self-check then offers targeted fixes, cheapest first:

1. Switch engines inside Athena. If your phone has more than one speech engine installed, switching is enough and downloads nothing. The switch applies only inside Athena and does not change your system default.

2. Download the missing voice data. The self-check can take you straight to the system's voice-data installer.

3. Open the system speech settings to check the default engine and language packs.

4. Set up cloud voices (see below).

5. If none of that helps, you can look for a third-party speech engine in your app store (MultiTTS, for example) and install it as your system engine. Athena provides no download link and takes no responsibility for the source or safety of third-party apps; please judge for yourself.

Cloud voices: bring your own API key

This is an optional advanced feature that fills the gaps your system engine cannot cover, in devices or in languages. It is never made the default, and Athena falls back to the system engine automatically if synthesis fails or you lose connectivity — basic listening always works offline.

Two providers are supported. You only enter your own API key under Settings → Listening → Cloud voices:

- Volcano Engine (Doubao): directly reachable from mainland China with no extra network setup. The full official 2.0 voice catalog is built in, covering 16 languages.

- Microsoft Azure Speech: for everywhere else. The voice list is fetched with your key from the full official catalog, covering hundreds of languages. Besides the key you must enter the region you chose when creating the resource (for example eastasia).

About voices and languages: we present everything the two providers offer, grouped by language and searchable. Pick a voice in the language of what you are reading — this has nothing to do with which languages Athena's interface supports. Before you register or pay, check the official docs below to confirm your language is on that provider's list. One Azure rule is worth knowing up front: if the voice's language does not match the text, nothing is spoken but the request is still billed. Picking the wrong voice raises no error — you simply get silence and a charge.

About cost: you pay the provider directly. Athena neither collects nor marks up anything, and sets no cap. Note that the two bill differently — Azure officially counts each Chinese character as two characters, so Chinese text costs twice the listed rate, while Volcano counts one Chinese character as one. The number of characters synthesized in the current session is shown in the app so you can reconcile it.

About security: your API key stays on this device only. It is never synced and never uploaded to Athena's servers. The cost is re-entering it on a new device; the benefit is that we never hold your third-party payment credentials. It is also excluded from system backup and device-to-device transfer so it cannot leak through a cloud backup.

Troubleshooting

"resource not granted" (Volcano): the key is valid but the service is not enabled, or the chosen voice has not been purchased in the console. Handle it under service management in the Volcano console. A valid key and a working synthesis are two different things.

Azure region error: the region must exactly match the one you chose when creating the Speech resource. For Azure China (21Vianet) resources, use a China region name such as chinaeast2.

Azure produces no sound at all: the voice's language most likely does not match the text. Azure stays silent in that case but still bills the request, so switch to a matching voice.

Speed slider does nothing: both providers support speed. If nothing changes, first confirm cloud voices are actually in use — Athena falls back to the system engine silently when offline.

The first sentence takes a moment: cloud synthesis needs one network round trip. Later sentences are pre-fetched, so playback becomes continuous.

Official documentation

Volcano Engine Doubao voice synthesis (voice list and billing): https://www.volcengine.com/docs/6561/1257544

Microsoft Azure Speech (language and voice support): https://learn.microsoft.com/azure/ai-services/speech-service/language-support