Voice features: dictation & read-aloud

Tholos AI lets you speak instead of type, and have answers read back to you — both running entirely on your machine. That matters because the built-in dictation in Windows and macOS sends your audio to the cloud; here, speech in and speech out never leave your device. There are two voice features: voice input (dictation) and read aloud.

These are part of the Professional and Business editions (and included in the free trial). Each uses a small speech model that’s offered for download the first time you use it.

Voice input (dictation)

Click the microphone in the chat input to start dictating. Your speech is transcribed into the text box as you talk — you’ll see provisional words appear and settle at natural pauses, with punctuation and capitalization added automatically. Click the mic again to stop, then edit and send (or keep going).

It’s useful for:

• Asking a quick question — dictate “summarize this document” or “what’s the term length?”, then send. Handy when you’re reading a paper file and your hands are busy.
• Long-form drafting — dictate into the chat box or the Writing Assistant instead of typing a long passage.
• Accessibility — a faster primary input method for anyone who finds typing slow.

The first time you use it, Tholos AI downloads a lightweight speech model. The language is auto-detected.

Read aloud

Turn on the read-aloud toggle in the chat header and completed AI responses are spoken aloud as they finish. Click it again while it’s speaking to stop playback. It’s good for reviewing a long answer away from the screen, for accessibility, or while multitasking.

The first time you enable it, Tholos AI downloads a small text-to-speech voice model.

Getting good results

• Pick the right microphone. Choose your input device in Settings if you have more than one.
• A clear signal helps. A decent mic and a quiet room improve dictation accuracy, just like any speech-to-text.
• Review before you send. Dictation lands in the text box so you can correct anything before submitting — speak, glance, tidy, send.