Speech & Reader

Open Settings → App → Speech & Reader to configure how Echo speaks — in the Reader, for spoken feedback, and for workflow speech.

System TTS

The Speech engine select offers:

• Automatic — uses the system engine. Kokoro is never selected automatically because it needs a downloaded model.
• System — the operating system's installed voices, always available when supported by your device.
• Kokoro (local neural) — an optional neural engine downloaded and processed on this device. The option is selectable before the download, because choosing it is how you start it; its label says whether the model still needs downloading, and Kokoro only speaks once the model is installed.

Below the engine, adjust the Playback speed (0.75x–2x), keep Follow spoken text enabled so reading stays on the active sentence, and confirm the setup with Test voice, which plays a short sample through the current engine and voice.

The Voices list shows the voices available to the selected engine. For the system engine it lists the higher-quality voices installed on this device with a language filter, and explains how to add more through the operating system's own voice management (on macOS: System Settings → Accessibility → Spoken Content → Manage Voices). For Kokoro it manages the engine download and the per-voice files, each about half a megabyte.

System speech is available on the supported macOS and Windows desktop paths. The Mac App Store build uses the same native speech service and does not need Accessibility or clipboard automation to speak text.

Workflow speech

The Text to Speech workflow action reads its interpolated input through the engine configured here. A workflow can leave the voice empty to use the configured default or provide a provider-qualified voice override. Speak automatic feedback extends this to eligible completed mode and workflow results; it is off by default and only speaks when the workflow has not already performed an explicit speech action.

Speech content stays on-device. Echo records bounded, content-free speech analytics only when usage analytics is enabled.

Reader

Reader opens in its own window from the popup or the tray menu and reads documents and public web pages aloud sentence by sentence through this same speech service. The full walkthrough — library, supported formats, playback dock, engine overrides, and resuming — lives on the Reader page.

Supported formats: TXT, Markdown, DOCX, and public http(s):// web pages. PDF is not supported yet — selecting a PDF file in the import dialog returns an error instead of importing it. Reader also requires selectable text; it does not perform OCR and cannot read a scanned or image-only page.

Library and resume: every import joins the library, sorted by when you last opened it. Reopening a document resumes at the exact sentence you left off on, even after closing Reader or restarting Echo.

Deletion: removing a document from the library deletes only Reader's own local copy and its reading progress. The original file or web page is never deleted or modified.

Reader's imported content, page text, and playback state are processed and stored locally; see Privacy & Data for where Echo keeps local files.