Let the AI agent run your presentation
9 October 2026
Snazzy Pro's assistant isn't only for writing slides. It's an agent with its hands on the same controls you use: it changes slides, starts and stops the recording, runs the prompter, and opens and runs your live class. You can type to it, or talk to it hands-free while you present.
In short: ask in the chat, or press ⌥⌘V for Voice Mode and say “next slide”, “start recording” or anything else. It answers out loud with your Mac's own voices and mutes your mic while it speaks, so its voice never ends up in your video. Going live, streaming and anything that can't be undone wait for your OK on screen.
Example 1: a sales demo, recorded hands-free
Priya records a short pitch for a prospect. She opens her deck, turns on Voice Mode, and never touches the keyboard:
- “Go to the first slide.” Then “Start recording.” After a short countdown, it's recording.
- She talks to the prospect: “Good morning, thanks for joining…”. While recording, Voice Mode only acts on lines that start with “Snazzy”, so this goes into the video and nowhere else.
- “Snazzy, next slide.” The slide changes instantly; simple commands don't wait for the AI.
- “Snazzy, in one sentence, what's the key point of this slide?” The agent knows which slide is showing and what its notes say, and answers out loud. Her mic is muted while it speaks, so the answer is just for her.
- “Snazzy, stop recording.” The video is saved, with a chapter for every slide.
Example 2: a teacher runs a live lesson from the chat
Mr Okafor wants his class to brainstorm saving energy. He types, one request at a time:
- “Start a live room for my class to brainstorm ‘How could our school use less energy?’” A sheet asks him to confirm what's shared; he clicks Start. Students join from the QR code in their browsers.
- “Post a short welcome message to the students.” It appears as a banner above the board on every student's screen.
- “Add two starter ideas of your own to the board.” Presenter ideas get a badge; students add theirs and vote.
- “Tell the class that voting is closed.” Then: “Create a new presentation project with 3 slides from the top ideas on the board, with speaker notes.” The class's ideas become next lesson's deck.
What the agent can do
- Slides: next, previous, go to any slide, open the Present window; write and fix the deck itself.
- Recording: start, pause, resume and stop; pick the camera, mic and layout; re-make a recording with a new layout.
- Camera Prompter: put your notes or a script under the camera, set the speed, start and pause it.
- Live room: open and close it, post messages, add ideas, close voting, read the board and turn it into a deck.
- Streaming: start and stop going live on YouTube, LinkedIn, Twitch and others (always with your OK).
- Setup: presets, settings and which AI model does what.
You stay in charge
- It works alongside you. Every control is still a button. Click whenever you like; the agent sees the change.
- It asks first. Opening a live room, going live online, deleting files and other things that can't be undone show a sheet on the window, and Voice Mode says “I need your OK on screen first.” The same goes for outside AI agents connected over MCP.
- You see every step. Each action it takes appears in the chat, including the ones Voice Mode ran.
- It doesn't need the cloud. Use Apple's on-device model or Ollama on your Mac, and nothing leaves it. Voice commands like “next slide” and “start recording” work even with no AI set up.
Setting up Voice Mode
- Press ⌥⌘V, or click the waveform button next to the mic in the chat. The first time, allow the microphone and speech recognition. Speech is recognised on your Mac when your language supports it.
- Say a command or ask anything. A bar above the chat box shows what it heard, and when it's thinking or speaking. Stop Talking cuts a long answer short.
- In Settings › Chat, choose the voice and speed, turn spoken replies on or off, and change the wake word. For a more natural voice, download an Enhanced or Premium one in System Settings › Accessibility › Spoken Content.
Commands that run instantly: “next slide”, “previous slide”, “go to slide 4”, “start recording”, “pause”, “resume”, “stop recording”, “start the prompter”, “stop listening”.
Tips
- Using a local model? Ask for one thing at a time, as in the classroom example. Smaller models are much more reliable that way.
- Want the agent's voice in the video? It's left out on purpose. If you want narration, put a script on the Camera Prompter and read it yourself.
- Prefer reading on screen? The chat text size is in Settings › Chat, or press ⌘+ and ⌘− in the chat.
For developers
Everything the agent can do is also an MCP tool. Turn on Snazzy Pro's MCP server in Settings › MCP (it only listens on your Mac, with a token) and Claude Code, Cursor or your own scripts can drive the same slides, recording and live room. The Claude Code plugin is in the integrations folder. Read-only tools are marked as such, and anything that can't be undone still waits for the person at the Mac to approve it.