I Got Tired of Dictation That Stops at Words
Most voice features on a computer are really just dictation with a marketing budget. You speak, text appears, and then you still have to use your hands to do anything with it. That's fine for writing emails slowly, but it's not actually controlling your computer.
LacPointer's voice control is different. Press Cmd+Shift+V on Mac (or the equivalent on Windows), speak what you want done, and it acts on your machine — opens apps, moves through UI, types into fields, controls playback. It goes through the real accessibility APIs that assistive technology uses, not some screen-scraping hack.
I've been using this for a few months now and there are specific workflows where it saves me a noticeable amount of time. Let me walk through them.
How to Trigger It
First, open LacPointer with Option+Space if it isn't already visible. Then press Cmd+Shift+V to activate voice mode. You'll see the bar shift into listening state — there's a visual indicator so you know it's actually hearing you.
Speak your instruction in plain English. Don't worry about phrasing it like a command. "Open Spotify and play my liked songs" works. "Slack message to the team channel saying I'll be late" works. You don't have to memorize syntax.
If you have a Persona active, the voice session runs through that persona's voice and system prompt automatically. So if you're in Coding mode, the assistant responds like a senior engineer. If you're in Stoic mode, it's calmer. The voice that speaks back changes too — each persona has its own assigned voice (alloy, verse, ash, sage, coral, or cedar).
What It Can Actually Do
Open Apps
Say "open Figma" and Figma opens. Say "switch to Terminal" and it brings Terminal to the front. This sounds trivial but when my hands are occupied — I eat lunch at my desk more than I should — it's genuinely useful. No reaching for the mouse, no Cmd+Tab hunting.
Click UI Elements
This is the part that surprised me most when I first tried it. LacPointer uses the accessibility APIs to read the button labels and interactive elements on screen, so when you say "click the Submit button" or "click Approve on the pull request," it finds and clicks the right element. It's not pixel-matching or taking a screenshot and guessing — it's reading the actual accessibility tree.
I use this most when I'm reviewing things. "Click Accept" on a diff, "dismiss" a notification, that kind of thing. Keeps my hands on the keyboard where they belong.
Type Into Fields
Say "type my standup update into the message box" and it'll put text into whatever field is focused or find the right one on screen. Combined with a Persona that knows your writing style, you can rough-draft messages out loud faster than typing them.
Control Media
Pause, play, skip, volume up, volume down — all work by voice. I know your keyboard probably has media keys. Mine does too. But when I'm in the middle of explaining something to a colleague and I need to kill the music, saying "pause" without breaking my train of thought is faster than finding the right key.
The Accessibility API Angle Matters
I want to spend a second on this because it's not just a technical detail. A lot of automation tools on macOS fake it — they take screenshots, use image recognition to find where a button is, then move the mouse to those coordinates and click. That approach is brittle. Button moves slightly, layout shifts on a different screen resolution, dark mode changes contrast — and the whole thing breaks.
Going through the accessibility API means LacPointer is reading what the OS itself knows about UI elements. The labels, the roles, the states. It's the same layer that screen readers use. That makes it more reliable across different setups and less likely to break when an app updates its layout.
On Windows it works through the equivalent accessibility stack (UI Automation). So it's not a Mac-only trick.
Pairing Voice With Personas for Different Modes
One thing I do that took a while to figure out is set up different Personas for different kinds of work, and then switch Personas before a voice session depending on what I'm doing.
For example, I cloned the built-in Coding persona and tweaked its system prompt to know my project structure and coding conventions. When I'm deep in a coding session and want to dictate something — "open the routes file in VS Code," "type a comment above this function explaining what it does" — the Coding persona handles it with the right context.
When I'm writing or doing admin work I switch to a different persona that knows my writing style, so when I say "type a reply to this Slack message," it sounds like me.
You can build and switch Personas from the LacPointer bar itself. Tap the persona icon, pick or create one, and every voice session from that point runs through it.
When It Doesn't Work Great
I'll be honest about the rough edges. Apps that draw their own UI without using system controls — some Electron apps do this — don't expose a proper accessibility tree, so LacPointer can't click their buttons by label. It sees the app but can't navigate inside it as precisely. Spotify's desktop app has some of this. I still use voice to open it and control playback through media controls, but asking it to click specific playlist items doesn't always land.
Also, very fast multi-step sequences sometimes need a small pause between instructions. "Open Notion, create a new page, title it Sprint Notes, and type the agenda" can work, but if the app is slow to load, the follow-on steps might fire before it's ready. Breaking it into two instructions is more reliable.
Neither of these is a dealbreaker. They're just worth knowing so you don't set your expectations to "it runs my entire computer with my eyes closed."
A Real Workflow I Use Daily
Here's a concrete one. Every morning I need to check a few things before I start coding: GitHub notifications, my calendar for the day, any Slack messages that came in overnight.
I press Cmd+Shift+V and say: "Check my GitHub notifications, read my calendar for today, and summarize any unread Slack messages."
LacPointer pulls all three through its integrations — GitHub, Google Calendar, Slack — and reads the summary back to me out loud while I'm making coffee. I don't touch the laptop for that first five minutes. By the time I sit down I already know what's waiting for me.
That's the kind of thing that sounds small until you've done it every day for a week and realized how much mental overhead it was cutting out of your morning.
Try It Today
If you already have LacPointer installed, you might not have used voice control at all — a lot of people discover Option+Space and live in text mode. Worth trying Cmd+Shift+V for five minutes to see what clicks for you.
If you don't have it yet, grab it at lacai.io/download. Free tier is real — 4,000 tokens every 15 hours, no credit card. Voice control is included.
One practical tip to start with: just try "open [any app name]" first. It's the simplest command and it'll tell you immediately whether voice mode is working on your setup. From there you can go as deep as you want.