VoiceType

I always hated voice dictation because it can't understand my German accent. It doesn't handle slang or domain specific terms very well and it makes a lot of mistakes. So in the end you always have to go back and make corrections and instead of making you more productive you waste more time. But AI has changed the voice dictation game because the big strength of AI is that it understands context. This way it can infer the meaning of words which makes it much more accurate. It can also correct mistakes for you and it uses punctuation intelligently instead of just guessing where a sentence is over. Today we will take a look at VoiceType, an AI voice dictation tool for Windows and Mac. We will compare it to the inbuilt Windows voice dictation tool and see if it's better. And we will try it out in different apps like an email client and a coding environment and see if it can handle voice dictation in these different situations and if we actually save time or if we still have to go back and make corrections all the time. And at the end of the video, I will show you how you can use special settings like a custom prompt to increase your productivity even further. My name is Florian Walther and this is the AI tool corner where I review the latest AI software to find out which ones can actually improve our lives and businesses. If you want to try it out yourself, I will put a link into the video description.

VoiceType is available for Windows and Mac as far as I know. And when we click on start free trial, we have to go through a short onboarding process. What you select here doesn't really matter. Just go through these few steps. And at the end, you can create your account and download VoiceType and install it on your computer. I assume that you know how to install a program on your computer. And when you then start VoiceType, this is what you see in the home tab. You can add words to your dictionary, which teaches VoiceType how to spell them. And on the left side, we see a history of our past transcriptions, and we can copy them if we need them. Again, in the settings, you can configure the hotkey you have to press to activate VoiceType. By default, this is set to the control and Windows key, but I use this combination to switch between different desktops. So you can change this to a combination that you don't use for anything else. I set this to alt and plus on the numpad, but what you use here is up to you. We have to hold this combination down to dictate, but there is also a hands-free mode where you press the hotkey once, then you can speak, and then you press it again to finish it. There are also a few display settings. Do you want to launch the program when you start Windows? I keep this on, and I just keep the default settings here. I also keep the output language to autodetect because I sometimes speak in English, sometimes in German. And this works really well at detecting what language I'm speaking in. And I can even mix German and English words and it still works. You can also set the microphone it should listen to. And I actually don't use this mic. You can see here I use a very cheap $20 Bluetooth headset with a bad mic, but it still works flawlessly for dictation. It's this one here. This is just a cheap Bluetooth headset from Amazon. But for recording this video, let me set this to my high quality mic.

The coolest feature is the custom system prompt, but we will try this out later. For now, let's just try out regular voice dictation. Let's try out VoiceType in Gmail. And it works in any software because it will put the text where your cursor is. Let's give it a try. And I will on purpose add some uhs and stutters and pauses and little mistakes into the dictation just so you can see how VoiceType corrects them automatically. Okay, so I press the keyboard combination and then I start dictating. Hey Florian, thanks for reaching out. Yeah, I'm going to make a video about VoiceType soon. Um, here are three things I uh really like about it. One, it's very accurate. Uh, two, it's customizable. Three, it auto detects languages. Thanks for subscribing to the channel. Regards, Florian. And after releasing the hotkey, it adds this text. And you can immediately see that it's formatted. It removed all the uhs and stutters and little mistakes I made. And it also changed gonna to going to to make it sound more formal. Now, if you prefer less formal spelling, if you prefer to keep gonna, I will show you how to customize this later. But this is impeccable spelling and grammar. And I really like it. But what is even cooler, it can also handle slang. So, for example, let's say I want to DM one of these random gym bros and I want to tell him, "Yo, you're looking hella sick, bro." You can see that it uses slang like hella and bro. And this is what I mean. The AI understands the context and it can infer the proper spelling of words from the meaning of the sentence. So here it knows that this is not a formal thing to say. So it doesn't use formal language. And this is really the strength of AI over older voice dictation tools. And it also auto detects the language. So let's say I want to speak in German instead. Here it recognizes that I'm speaking German and it enters the correct German text with proper grammar and you can even mix different languages which I sometimes do when I dictate my video notes.

We can even use it for vibe coding. So here I am in Cursor which is a development environment and I tell it please go into the mutations.ts file and call the useCompleteOnboarding function. And this is super cool. A regular voice dictation tool would have never spelled this correctly because it had to figure out that this is a file name with a period and a file extension and useCompleteOnboarding has to be spelled in camelCase where all the words are together because this is how you spell function names. But the AI understood from the context of the sentence that this is a function name so it spelled it correctly. Now, if you dictate something but you forget to put your cursor in the correct position and the text is not applied, you can either dictate it again or you can go to your history and copy it from here.

So, AI is really good at detecting speech, but there are still a few problems. What if you dislike this style? What if you don't want this very formal spelling in emails? Or if you want less punctuation because you write on a website like Twitter and you don't want to end each sentence with a period because it doesn't look natural anymore. If we can't customize this behavior, the tool becomes less useful and we waste a lot of time again. But in VoiceType, we can actually customize this behavior with the custom system prompt. This is really cool. This is just an AI prompt that will be inserted before your dictation is generated. Here we can give the AI instructions and we can even use VoiceType to create this instruction. So when I dictate only a single sentence, don't add a period at the end and don't capitalize the first letter of the sentence. There we go. We have to save the changes. And now when I say yo, you're looking hella sick, bro. It doesn't turn it into a formal sentence anymore. It doesn't add a period at the end and it doesn't capitalize the first letter. Just like I instructed it. But when I say multiple sentences, it will still write them out properly. So for example, sir, you look really good. I admire your strength and performance in the gym. Now it uses formal spelling again because we told it to only use this style when I dictate a single sentence. And you can add multiple instructions here. For example, I told it to avoid em dashes. And when my dictation starts with Hey VoiceType, treat everything that follows as instructions rather than verbatim dictation. So now I can do something like this. Hey VoiceType create a grocery list and make each item start with a fitting emoji. Put the following items onto this grocery list. Um cheese meat milk bread. So now because I started the dictation with Hey VoiceType, it did exactly that. It treated everything that I said as instructions rather than direct dictation and it created this grocery list. So you can do some really cool stuff with the system prompt. And if you want less formal spelling, for example, if you want to let it write gonna instead of going to, you can add this here as another instruction. What you enter here is completely up to you.

Now, let's compare VoiceType to Windows's inbuilt voice dictation tool, which we can activate when we press the Windows key and H. I will now try this out, and I will dictate the same email with the same mistakes that we dictated to VoiceType earlier. So, Windows H starts it. Hey, Florian, thanks for reaching out. Yeah, I'm going to make a video about VoiceType soon. Um, here are three things I uh uh really like about it. One, it's very accurate. Two, it's uh customizable. Three, it auto detects languages. Thanks for subscribing to the channel. Regards, Florian. We stop it by pressing the hotkey again. And as you can already see, this looks terrible. There are tons of mistakes in here. No formatting and you can't customize it in any way. So, I can't use this. VoiceType, on the other hand, gets almost everything I say right, even with my bad accent. And I have to say, using a good voice dictation tool really makes using a PC much more comfortable. For one, it's faster, but it's also less exhausting because you don't have to type all the damn time. And you are also less glued to the screen because, for example, sometimes I dictate something and I look out of the window while doing it instead of having to look at the screen all the time. And this makes it much more enjoyable. If you want to try out VoiceType yourself, I will put the link into the video description. They have a free trial. Subscribe to the channel for more AI tool reviews. And then I hope I see you in the next video. Have a nice day. Take care.