Voice Input
Voice input lets you speak your prompt instead of typing it. Useful on mobile, for accessibility, and for capturing rough ideas quickly before refining them.
Where it’s available
Section titled “Where it’s available”- AI Course Generation chat
- AI Lesson Generation dialog (from any course’s overview)
In both places you’ll see a microphone icon in the input toolbar, next to the attach and expand buttons.
How to use it
Section titled “How to use it”- Click or tap the microphone icon.
- Allow microphone access if your browser asks.
- Speak your prompt. A live waveform shows your audio level, and your transcript appears below it as you speak.
- Choose how to finish:
- Insert (checkmark) - drops the transcript into the message box so you can review and edit before sending.
- Send (paper plane) - sends the message immediately, skipping review.
- Cancel (X) - discards the recording without inserting anything.
You’ll hear a soft tone when recording starts and stops, so you can confirm the microphone is active without needing to look at the screen.
Live transcript
Section titled “Live transcript”While you speak, two kinds of text appear in the transcript area:
- Solid text - words the recognizer has finalized. These are locked in.
- Italic muted text - words still being interpreted. These may change slightly as you keep speaking.
The transcript scrolls automatically to keep your latest words visible. If you scroll up to re-read something, auto-scroll pauses until you scroll back down, so you can review what was captured without fighting the scroll.
Recording length
Section titled “Recording length”Recordings auto-stop after about 60 seconds of continuous activity (a browser limitation). Around 50 seconds, the elapsed timer turns amber and shows a countdown so you know to wrap up.
For longer prompts, finish the current recording, then start another. Your previous text stays in the message box so each new recording appends to it.
Combining voice and typing
Section titled “Combining voice and typing”You can mix voice and typed input freely. For example:
- Type the start of your prompt, then dictate the rest.
- Dictate a rough idea, then edit specific words.
- Dictate one sentence, send it, then dictate a follow-up.
Voice transcripts always append to whatever’s in the message box. They never replace existing text.
Browser support
Section titled “Browser support”Voice input uses your browser’s built-in speech recognition, so support depends on the browser:
| Browser | Voice input |
|---|---|
| Chrome (desktop and Android) | Supported |
| Edge | Supported |
| Safari (desktop and iOS) | Supported |
| Firefox | Not supported (no microphone button shown) |
The microphone button is hidden entirely in unsupported browsers, so you’ll never see a broken control.
In Chrome and Edge, speech recognition runs through Google’s service, which means it requires an internet connection. If you go offline mid-recording, you’ll see a brief error message.
Privacy
Section titled “Privacy”Audio is processed by your browser’s speech recognition service (Google in Chrome and Edge, Apple in Safari). Slate does not store your audio recordings or send them to its own servers. Only the resulting text transcript is used, and it stays in the chat conversation like any typed message.
- Speak clearly and at a normal pace. Short pauses help the recognizer commit words faster.
- For names, technical terms, or unusual spellings, use Insert instead of Send so you can fix them before sending.
- Background noise reduces accuracy. A quieter environment (or a headset microphone) noticeably improves results.