It would be a stretch to say that anyone really misses Cortana on Windows 11, though I'd argue it was at least a much more interesting addition than Copilot currently is. The idea of a proper voice assistant on a PC did have potential, it was just severely underutilized with Cortana, and even more so with Copilot. But what if you could add voice commands to Windows 11?
That's what Talk2Windows is all about. This app leverages voice recognition to make your PC controllable using just your voice, and while it's not perfect, it works surprisingly well, doing many things you could never do with Cortana. It even blends in some elements of Voice access, giving you basic controls over certain OS and app features, too. Let's talk about it.
On this day 10 years ago, Cortana landed on Windows Phone as a digital assistant
Sadly, Cortana gathered a lot of flak following its release on Windows 10. In the end, Microsoft pulled the plug on the virtual assistant in 2023
Setting up isn't too hard
Though it requires some available space
In order to use Talk2Windows, you'll need to install a program called Serenade, which is designed to help developers code using their voice in supported IDEs. This is a relatively large install since it features voice recognition models, but it's fairly easy to set up otherwise. The initial process is a bit confusing as you're given links to install plugins, but in reality, clicking the links only opens a web page, and after that you can dismiss the plugins screen to get through and actually use the app.
The only other thing you need is to download the Talk2Windows repository from GitHub, extract the archive, and run the setup.ps1 script in the files. This will add the voice recognition commands to Serenade's library, making it so your PC can actually respond to the things you say. To run the script, you'll want to launch Windows Terminal/PowerShell with administrator privileges beforehand, navigate to the folder, and run the file that way. You may also need to use Set-ExecutionPolicy -Bypass in order to be able to run the script.
Once it's done, all you need to do is set Serenade to listening mode, and you're off to the races.
6 PowerShell scripts to automate and speed up your workflow
When you work through the CLI you save a lot of time. Here are PowerShell scripts to speed your workflow.
A surprisingly comprehensive library of commands
There's a ton you can do
Going into this, I really expected Talk2Windows not to be much more than a gimmick with a couple of recognized phrases that don't do much at all. I was almost disappointed to find out I was very much wrong in my assumption, as there's quite a lot available here. You can ask your PC to open apps and websites, ask about the weather, and even more technical details like the current CPU temperature on your PC. The list of phrases is genuinely very long, and it even includes some more direct commands like closing a tab. The best part is this isn't even browser-dependent, as I tested it with Vivaldi and it worked.
Of course, every command has to programmed manually, so there's a finite number of things that can be done, but the developers have enabled support for quite a lot. In terms of apps, you can open things like 7-Zip, CrystalDiskMark, Discord, Firefox, or VLC, among many others, and you can just as easily close said app. It even supports installing some apps from the Microsoft Store, including all the ones mentioned above, but also things like Opera GX, Visual Studio Code, and Spotify.
For websites, a lot of them are also supported, including Amazon, Apple, CNN, Discord, PayPal, Wikipedia, and so on. And then there's the "check" command, which can get information from both your PC and the web. You can ask about CPU temperatures, internet speed, ping latency, and more. Some shortcuts like closing tabs, scrolling, or zooming are also available, you can insert specific bits of text or special characters, and so on.
You can even get some fun out of it by asking your PC to tell you a joke or give you a random quote. It blends usefulness and fun, which is exactly what I would have wanted Cortana to be. In the beginning of its life, it felt like Cortana only really had the fun element, and then Copilot seems to only care for the more useful actions, and even then, it's not great for that.
Home Assistant's Piper just got a big update for local voice assistants, but you'll have to wait to use it
This will be a massive update for people who run a local LLM paired with Home Assistant.
A more private solution
No online processing needed
One great aspect of this solution compared to Microsoft's is that none of your voice data needs to be sent to the internet. You can set Serenade to run using a local server so all the processing is done on device rather than using the internet. You can also make it so that no voice data is sent to Serenade to help improve audio recognition in the future, if you want to be extra careful with your privacy.
Compared to using Copilot and Cortana, that privacy benefit is a very welcome advantage over other voice recognition solutions. Plus, none of that data is sent to Microsoft in the first place, which depending on how much you trust the company, could still be a benefit.
Stop using these 4 Windows features if you care about privacy
Windows can be a solid OS, but you should be leary about using these features.
It's a little stiff
Natural speech? Not here
The main problem with the way Talk2Windows works is that it requires you to speak in a sort of stiff way. It's not really programmed to handle natural language, so you have to speak the commands exactly in the format they're designed.
For example, when I wanted to check the CPU temperature, I had to say "Check CPU temperature", not "Check the CPU temperature". Any phrases where you might add words like that for a more natural flow have to be trimmed to this sort of roboticized format. Of course, I'm sure this could be programmed to also recognize sentences with those words, but all that work has to be done manually, and I'm sure it's not the biggest priority for the developers compared to just adding new features. I would love to see it, though.
What could be interesting would be the ability to create your own phrases or commands that can be recognized. Maybe not the entire phrase, but say you want to add a website to the list of supported websites — it would be cool if you could edit a file to add the name of a website that would be recognized for that command. It might be easier for the developers to make a tool that allows you to modify the recognized commands yourself than to add all those commands themselves, so that would be cool to see.
This free Obsidian plugin turns my voice into notes, and it all runs on my computer
Using Whisper plugin with its local LLM, I use Obsidian to transcribe my voice notes and audio files to text on my computer.
It's not for everyone, but it's really cool
Even the greatest voice assistant in the world probably wouldn't please every Windows user, so I know not everyone is going to be interested in something like Talk2Windows. However, I do think this is one of the best implementations of voice controls yet in Windows, integrating so many features that have previously been part of separate solutions. It has a little bit of voice access and a little bit of Cortana, and the benefit of privacy on top of all that.
