Live Voice Course Planning
Live voice lets you plan a new course by speaking with Slate AI in real time. The conversation captures your course topic, audience, objectives, length, and tone as you talk, and at the end you decide whether to attach a reference document and build the draft.
When to use it
Section titled “When to use it”Live voice is built for one job: planning a new course quickly. It works best when you have a rough idea in your head and want to talk it through, the same way you might brief a teammate. The whole conversation is usually 5 to 10 minutes.
For dictating into the typed AI chat, use Voice Input instead.
Two ways to start
Section titled “Two ways to start”- From the dashboard: Click the voice icon next to the + New course button. Each click starts a fresh course conversation, even if you have other chats open.
- From inside an AI Chat: Click the live voice icon in the chat toolbar. This continues the current chat with voice, picking up the requirements and recent messages so you don’t have to recap.
During the conversation
Section titled “During the conversation”A central orb fills the screen and reacts to who is speaking. Slate’s voice greets you and asks about your course; you answer naturally, and Slate captures the requirements as you go. You can speak as little or as much as you like. Short, conversational answers work better than long monologues.
You’ll see a small readout in the corner of the screen showing how many minutes you have left this month. The conversation itself runs up to 10 minutes; if you need more, the orb will let you know it’s wrapping up and you can start a new one.
When Slate has enough to work with, it’ll summarise what you’ve shared and ask you to confirm. Once you say yes, the conversation ends automatically and the screen moves on to the build step.
Optional reference material
Section titled “Optional reference material”Before building the course, you’ll see an Add reference option below the Build my course button. You can attach a single document (PDF, slides, Word, plain text, or Markdown, up to 20 MB) and Slate will use it as source material when generating the draft. This is the same source-material pipeline used in the typed chat.
The Build my course button is briefly disabled while Slate reads the file. When that’s done, click Build my course to generate the course and open it in the editor.
If you’d rather skip the attachment, just click Build my course without adding a file.
Picking your voice and orb
Section titled “Picking your voice and orb”You can change the voice that speaks to you, the colour of the orb, and the geometry of the orb in Settings > Slate AI. The previews on that page let you hear each voice and see the orb in any combination of colour and design before you choose.
Five orb designs are available:
- Slate - rounded squares, a quiet nod to the Slate brand mark (default)
- Waves - three flowing waveforms
- Floral - five-petal bloom, slow and steady
- Minimal - just the orb, clean and quiet
- Pulse - concentric rings rippling outward
These choices save automatically and apply the next time you open the live voice screen.
Language
Section titled “Language”Live voice speaks in your Default source language (set in Settings > Profile). The voice greets and stays in that language from the first word, and the generated course is drafted in the same language.
Before voice starts, you can change the course language with the Languages chip in the top-right of the live voice screen during the review step (the moment after the conversation ends, before you click Build my course). Picking a language here also updates your Default source language so the next session starts in the same language.
Switching language partway through a live conversation is not supported. If you ask the agent to switch mid-call, it will let you know it can’t, and tell you to end the call, change the language from the AI Chat input or Settings > Profile, and start a new session.
Monthly minutes by plan
Section titled “Monthly minutes by plan”| Plan | Monthly voice minutes |
|---|---|
| Free | Not included |
| Standard | 30 minutes |
| Pro | 90 minutes |
A single conversation runs up to 10 minutes. Standard plans cover roughly three to six course conversations a month, depending on how long you talk; Pro plans cover nine to eighteen.
Team voice pooling
Section titled “Team voice pooling”Team workspaces share a single pool of voice minutes instead of giving each member their own allowance. The pool is 90 minutes per seat, so a 3-seat team has 270 minutes a month, a 5-seat team has 450, and so on. Every team member draws from the same pool, including team members whose personal plan is Free or Standard.
The pool refills on the team’s billing cycle. Seat changes update the monthly pool immediately, with billing prorated automatically.
When you’re working in a team workspace, the voice usage card in Settings > Slate AI and the readout on the live voice screen show the shared pool (“Team: 240 of 450 minutes used this cycle”) instead of the per-user “X of 30 minutes” cap. Your personal workspace allowance is unaffected: it stays on whatever your personal plan provides.
Free plan
Section titled “Free plan”Free accounts can preview every voice and orb option in Settings > Slate AI, including playing samples of each voice. The actual live conversation is a paid feature and requires Standard or higher.
When you click the live voice button on a free account, you’ll see an upgrade screen with the orb at full size, the voice and orb pickers, and your default course language confirmed before you commit. Any choices you make there save to your profile, so they’re already in place the moment you upgrade.
Privacy
Section titled “Privacy”Audio is streamed directly from your browser to Google’s Gemini Live API and never passes through Slate’s servers. Slate’s role is to issue a short-lived access token; the conversation itself runs over a WebSocket between your browser and Google.
Slate stores the resulting transcript with the chat (so you can scroll back and read what was said), the structured course requirements captured from the conversation, and the total minutes used against your monthly allowance. Slate does not store your audio recordings.
Google’s handling of the audio is governed by the paid Gemini API tier: per Google’s published terms, paid-tier inputs are not used to train Google’s models. Google may log inputs and outputs transiently for abuse monitoring (currently up to 55 days per Google’s documentation).
For the full privacy and legal picture, see Slate’s Privacy Policy (which lists Google as a sub-processor for the Live API) and Google’s Gemini API Additional Terms of Service.
- Have a rough idea in mind before you start. “I want to teach new managers how to give feedback” is enough to get going.
- Speak conversationally. Don’t worry about pausing to think; the orb listens patiently.
- If you want a reference document used, mention it during the call. Slate will let you know you can attach it before the build step.
- Use Keep talking on the build screen if you remembered something else and want to add it before generating.