Transcription
These AI voice apps for your Mac could potentially save you more time than all your other tools combined. To start, they let you type four times faster than the average person with very accurate dictation that completely puts Apple's to shame. And if you think about how much time you spend typing on your computer, that's already pretty ridiculous. I mean, just imagine all your emails, notes, messages, and AI chats being done at the speed of thought.
But it's actually just the beginning of what these apps can do. All three come with their own unique set of AI features aimed to supercharge the human voice, range from free to pricey, beginner to advanced, so there's something here for everybody. Let's put them head-to-head, so by the end of this video, you know exactly which one is for you and how to use it.
Kicking us off is WhisperFlow. This one is the new kit on the block, but boy, is she coming in hot because it legitimately blew my mind the first time I used it. And I chose it first because it's not only really impressive but very simple. Let's reply to Caitlyn's email here using WhisperFlow and see just how fast we can do it. Go ahead and start the timer. She asked, "What is the big deal with WhisperFlow?" I'm going to go ahead and start it in hands-free mode.
"Hey Caitlyn, great question. WhisperFlow is one of the best dictation apps I've ever used." I'm going to stop it there, actually. Check this out: One, it's already formatting it correctly as an email. There's an exclamation point, which is maybe a little weird, but I don't mind it. And then it's literally got her name spelled correctly as well, because it can see limited text that's on the active application. Pretty sweet, but we're going to keep going.
WhisperFlow can do all of the following: One, it dictates really, really fast. Two, it's able to, um, fix, uh, filter, I mean, filler words. Three, it handles punctuation completely on its own. Four, it adapts to the app that you're in, using proper punctuation in emails but keeping things more casual in messages. This also helps it get names right the first time. Five, it generates a dictionary of words that are personal to you, allowing me to say something like "Tai teaches Tech," knowing it'll be spelled correctly. And six, it has whisper and command mode, but I'm going to show you that in a second. Anyway, hope this helps. Let me know if you have questions. Best, Tyler.
Okay, this is a very complicated, uh, thing for AI to get, but let's see how it does. Okay, ignore the, uh, change up here. I'm pretty sure that's an Apple thing because that doesn't happen anywhere else. But we see it dictates really, really fast. It is automatically getting the punctuation completely correct with the numbering, which is really cool. It completely fixed this sentence if you remember me messing up a lot when saying it. And then it adapts to the app you're in, like we saw up here. This "Caitlyn" with a new line as a classic greeting is automatically getting added by Flow. And as we can see in number five, it did get "Tai teaches Tech" correctly, though I would have preferred to have a capital T here. But either way, this is amazing because we all have words that are very personal to us. WhisperFlow just makes this a very personal experience with the different spelling here.
The dictionary, by the way, can be accessed in the menu bar by clicking "Add word to dictionary." You can add the obvious ones manually off the bat, but Flow also learns them automatically as you use it and correct its output. For example, I didn't have to tell it how to spell my girlfriend's sister or a doctor's name. I just went in and corrected the false output, and Flow let me know it caught that and won't mess up again. Now, the other two apps do have dictionaries, but they work a little bit differently, and you do have to add words manually. If Flow gets this right and can add words consistently, accurately, automatically, all the time for people, then I think it's going to be huge for winning over a big user base.
But we're going to step it up even more, and I'm going to show you these two modes right here. Whisper mode is exactly how it sounds. Actually, if you find yourself around a bunch of noisy people and are still too lazy to type, don't worry. At the small cost of looking just a tad bit crazy in front of your peers, you can whisper into the microphone, and Flow will still sort it all out. Since I'm in social, I just tested this at home with my TV on, and it did work very well, but it also worked for super whisper and Mac whisper.
Now, command mode is where things get extra juicy, and there's two ways to use it. One, if you highlight your text, you can have Flow edit it. Just say, "Hey Flow," and then tell it what you want. For example, "Hey Flow, make this more concise," or "Hey Flow, translate to Spanish." This could be incredibly useful because you can simply turn on Flow and then stream of conscious reply to, let's say, a professional email. It'll look terrible when it's done, but then you just highlight and then say something like, "Hey Flow, turn this into a professional email while maintaining as much of the original wording as possible." And there we go. Wild stuff.
Command mode will also give us access to Google Search and Perplexity, and I think more integrations are on the way. And I do not think there is a faster way to Google. Check this out: If I have a question, I just have to hold down the function key and say, "Search Google for, is a tomato a fruit or a vegetable?" There we go. Instant answers. It's so simple, but so effective once you get in the habit of using it.
All in all, I was very impressed with their dictation. It has like that magical, like, it just, it just works feel to it, and it gets all these little things right that you don't realize are important until you start trying using dictation all the time like I did. It's fast enough where I can add even just a few words at a time and know it's going to be faster than typing. It lets me add in the middle of sentences without having to worry about punctuation being off. It formats and adds punctuation on its own very, very well. And for each app, I asked the team what's in store for the future. And for this one, they emphasized, "We're going to be able to have even more and more control over the output." For example, if you only text your friends in like all lowercase, you'll be able to set that up accordingly, or you'll be able to record like a 10-minute ramble inside of Flow and then have it organize it however you want, maybe like for a CEO PDF or something like that. This is the kind of feature that the next two apps will have as well. And finally, and probably most importantly for a ton of you, they're coming out with an iOS app, so you'll be able to get the same super fast dictation on your phone on the go. There's like a free trial if you want to test it out, but it's going to be like $12 to $15 a month if you really want to use this thing like you should.
But if you're a stickler for free, don't worry, MacWhisper has got your back. This, along with SuperWhisper, introduces a whole new set of features that supercharge your voice. In fact, I would be doing this app a disservice if I only talked about its free dictation feature. This app is a voice-to-text powerhouse, and a ton of people love it. I mean, feast your eyes, it's like an all-you-can-eat transcription buffet. It lets you work with OpenAI's free Whisper speech-to-text models to create and work with high-quality transcriptions of just about anything: video files, audio files, meetings, YouTube videos, audio playing from anywhere on your computer, you name it. And it's all done completely locally, so no data leaves your computer.
Being able to transcribe all this audio is an essential tool for content creators, students, academics, and business professionals. It makes you feel like the world is at your fingertips because you can take anything, transform it into text, and then feed it into AI to learn more about it. It's as simple as dragging the file into MacWhisper, hitting "Copy Transcript," and pasting it into your AI of choice. But the pro version, which by the way, is a one-time buy, unlocks quite a bit more. You'll be able to batch transcribe multiple files at once, use cloud transcriptions, which are going to be faster and very accurate for older Macs, transcribe YouTube videos straight from their URL, and most importantly, use your favorite AI service to chat with any transcript.
Now, I don't know if you've noticed, but most apps charge like $8 to $15 a month just to unlock the AI features. But the developer of MacWhisper is a real one. Instead, you can plug in what's called an API key, which means you're only going to pay for the AI services you use, and you get to pick which one it is. Now, I realized that might sound too technical or expensive, but don't worry, it really isn't. Setting up this API key is essentially as easy as going to the website linked directly within MacWhisper, adding a credit card, clicking one button to generate the key, and then pasting it into MacWhisper. Now, my account with an AI service and MacWhisper are perfectly connected, and anytime I use MacWhisper's AI features, it will use this API key to send the prompt to ChatGPT, and the account will get charged accordingly. Bookmarking this website will let you track and put limits on how much you spend, making sure you don't go over.
But to give you an idea of just how cheap the API can be, if you send over the entire Sorcerer's Stone to ChatGPT and then tell it to make a summary that's like 10% the size of the book, it's going to cost you a whopping 2 cents. But yeah, if it's still confusing and you want me to make a video on setting this up step-by-step, then I would be happy to. Just let me know in the comments.
Now, back to MacWhisper features. Transcriptions can be saved here on the left. For each, you can play the file and follow the transcript in real time, assign speakers to each line manually, which is going to be great if you're in production, and ask the transcript questions using a live library of fully customizable AI prompt presets. You can also translate into other languages and export into just about any file type you can imagine.
Now, MacWhisper gives us a free version of dictation that you can also supercharge with the pro version. It's actually a newer feature that's in beta at the time of this recording, but there are three key features that make it better than Apple dictation in my opinion. The first point, we already established, it's that OpenAI Whisper models are going to be way more accurate than Apple's built-in dictation. The second is that, similar to Flow, you can create a personal dictionary of commonly misspelled words, and you can even get fancy and add things that are just like impossible to say, like links or emails or addresses. So, for example, with this one right here, whenever I actually say "Instagram link," the actual link is going to get added. And again, while this is a very useful feature, it does not add words automatically like Flow, so keep that in mind.
But it does have this unique find and replace, which allows us to add links. So whichever one's good for you. And the third reason is the most powerful, though it's also paid. It's these AI prompts. These allow you to run everything you say through an AI prompt before it gets pasted into the computer. This is essential for being able to really accurately add, like, emojis or punctuation, because OpenAI's Whisper model does not do it by default, which is the one thing that Apple dictation does have over it. But you can also get really fancy with these AI prompts. For example, you can have it automatically translate what you say into Spanish before getting pasted into the app, or you can have it fix your grammar and make it sound very powerful, or maybe not even transcribe what you said at all. And instead, if you want to have the AI answer any question that you say out loud, you can have it do that as well. The beautiful part about this is it's really up to you and your creativity and ability to prompt. And changing the prompt your transcription gets run through is as simple as clicking this little sparkle icon that pops up when you start dictation.
For example, here, if I select "just dictate," then we're going to be using just the Whisper model, which is fast, but it doesn't do any fancy stuff like spoken punctuation. For example, if I want to say "exclamation point," it's literally going to say "exclamation point." However, this is a really easy fix because you can create a prompt that not only lets me dictate any punctuation or emoji I want, just like Apple dictation, but also I can add in that it automatically cleans up any filler words or mistakes I make while talking out loud. With this prompt, which by the way, I'm including everything in the description below, I was able to get surprisingly similar results to Flow. And to show you just how creative and useful these prompts can get, check out this "Hey Marge" example I created. It's designed so that I can give AI instructions to do anything I want at any point in the recording, as long as I start the instructions with "Hey Marge" and end with "Thanks Marge."
For example, I can activate MacWhisper and say something like, "Hey Marge, I'm about to think through everything I need to do today, but it might be all over the place, so I need you to make a concise to-do list when I'm done. Thanks Marge." And then continue to just rant out loud, freeform thought on what I need to do today. And then when I'm done, the result is not going to be my transcript, but instead the to-do list that I asked for. So this lets us really just take what we say and then transform it into the digital world and organize it instantly.
And finally, if you've made a library area of prompts that you use for different situations, you can set certain prompts to be defaulted based on what app you're on. If I'm in Mail, it'll automatically format it as a professional email, or if I'm in Messages, it'll just use a regular dictation so I can say whatever I want without it getting changed. But again, if you just want the basic dictation, you can always get MacWhisper and try it. But if you want the custom AI prompts that really, you know, give you the razzledazzle, then you're going to have to pay for the pro version as a one-time fee, and then you can use the API, which is not going to be that expensive. Once you get everything set up, it's an exceptional bang for your buck. Just make sure you select the cheap model like GPT-4o mini when you're adding your API key, though. If you need a little extra intelligence, you can always upgrade to give it a little extra boost. But before whipping out your wallet and clicking my affiliate link, please and thank you.
Let's talk about SuperWhisper. Like WhisperFlow, SuperWhisper is built around dictation. It has it at its core. The difference, however, is if Flow is like a plug-and-play kind of gal, then SuperWhisper is more of a tinker-and-tune kind of guy. In other words, if you like customizing your tools, you are going to love SuperWhisper. It takes those AI prompts that we just talked about with MacWhisper and then gives them a healthy dose of protein, and the things you can do are actually pretty mind-blowing.
But let's start with the basics. Let's hit the hotkey to start recording, and whoo-hoo, that is a, that is a big black box. Don't let it intimidate you, though. Just a little tickle right here in the corner will humble it right quick. This is how I prefer to use it, but we're going to actually open it back up again because this makes it a little easier to learn. We get this cool little waveform to show that it's recording, and we'll see down here we have "Voice" selected. This means it's acting like normal dictation. You can talk, and then watch your words fly on the screen. When you're done, par for the course.
But I can switch the mode as well, which does the same thing as switching the AI prompt in MacWhisper. For example, if I select "Email" right here, my transcript is run through an AI prompt that formats it with proper salutations, paragraphing, and sign-offs. If I select "Notes," the prompt is going to organize everything I said into bullet points with sections and headers. And as you can see, this is behaving a lot like MacWhisper. The difference, however, is MacWhisper only lets you feed your transcript through AI. SuperWhisper lets us include information from our clipboard, our active application, and our computer audio. These inputs, along with the right prompt, open up a world of possibilities for you, and the power is in your hands to create the modes that help you the most.
But for me, with great power came great confusion on what I'm even supposed to do with it. It does come with a handful of built-in modes, but they felt a little basic to me. And with all these inputs that we can add into AI in real time with our voice, I wanted to see how far I could push it. So I made these five modes that are actually changing my life. They're in the description if you want to check them out.
First, "Better Dictation." This is the same prompt that I used with MacWhisper for really clean, clean, flow-like dictation. However, there's one thing I was able to do here that MacWhisper cannot do, and that is reference the text in the active application to spell names correctly. This is basically the same functionality that WhisperFlow had when it got Caitlyn's name correctly in the email. It accounts for stutters, it'll add random bullet points or numbering, it'll do punctuation, emojis, etc. On my M1 Mac, it was about a half a second or one second slower than Flow, but what Flow can't do is this "Ask AI" mode.
So basically, I'm able to turn on SuperWhisper, select this mode, and now I can ask AI anything about my active application or what's on my clipboard or just anything at all, even if it's not related to the other two. It's my personal favorite because I basically can just ask AI anything at any point on my computer about anything.
"Hey there, future Tyler. I realized that I probably just gave you a lot of information that was not very specific, so I'm going to show you 'Ask AI' in SuperWhisper with this example. Let's go ahead and plan a trip to Japan. So we'll say 'Japan Travels,' and let's say I activate the 'Ask AI.' Let me bring this over here. Two monitors, so I have 'Ask AI' selected in SuperWhisper. I will start it and say, 'I'm going to Japan and I want to visit three cities that are not popular but worth going and underrated for tourist travels. I like the forest and nature in calm areas. Can you list three cities in Japan I should go to?'" And then I will hit enter. So basically, I'm just, you know, asking a question, and then it's literally going to output the answer because the prompt is basically allowing AI to discern, all right, just answer the question.
But I can also do something like copy this right now and then start it again. "Can you make a table of the cities mentioned in the clipboard and create an itinerary? So a row is either going to be morning, noon, or evening, and each column is going to be a city, and then within the cells, I want a specific place within each city I can go to in that time frame." See if it gets this. I butchered that, I apologize everyone. But here we go. So basically, it was able to, uh, I was able to copy this, and based on the prompt again, it knew to just reference the clipboard because I told it explicitly to check out the three cities in the clipboard. So we have morning, noon, evening, we have these two or these three cities. Perfect. And it gave specific places I can go in each of these time zones. So it's great.
And then another thing it can do is reference the active application. So I'm not even going to copy anything, and instead, I'm just going to ask, "I have never been to these cities before. In the active application, can you let me know anything I should look out for for each of these cities as a traveler who does not speak Japanese and has never been to Japan before?" And as we can see, it is literally just able to, you know, reference the text that is in this Notion document, or if I was in a Word document or anything like that, it could see it. And yeah, that's basically "Ask AI." So you really are just able to just hit the function key or whatever your activation key is and just go to town, tell AI what you want it to do, and it'll basically do it.
Number three is the "Meeting Summarizer." So you can select this mode within SuperWhisper, have it be hands-free, so we just have it recording the whole time, and then it will listen to any meeting on any app and at the end of it, it will provide a summary, next steps, and any open or pending items for next time. It can even detect different speakers, which is pretty impressive.
Next up, we have the "Scripting Mode." So sometimes talking out loud is just better than thinking inside your head or typing on a keyboard. And for me, this is especially true for scripting content because I want to feel relatable, and I do that better when I'm talking instead of typing. So with SuperWhisper, I can activate this mode, and it will basically remove all of my messes up, so I can just say the same line over and over again, iron it out, and then I basically tell the AI, "Hey, just ignore all of the previous takes, only keep the last best one," and I will output my entire script that I just said out loud and keep only the best takes. This has saved me hours.
And finally, we have "Rewrite Clipboard." If you ever have a hard time writing something, you'll be able to copy it to the clipboard, select this mode, and then ask AI to change it however you want. For example, I could say, "Make my clipboard more concise," or I can get crazy and say, "Create three succinct and friendly variations," and then it will output all three in bullet form, and I can use these outputs to inspire my writing.
"Hey, future Tyler here. Surprise! I actually made a sixth mode that I wanted to share with you guys because I found it too useful and a little bit unique from the others. It is a "Journal Mode." So it basically allows me to just journal with my voice as like a train of thought as opposed to, you know, writing things down in a notebook, which I'm, you know, it's not really for me. But basically, what I would do, actually, I'm going to include a little tip here. If I hit Command Shift K whenever SuperWhisper is open, I'll be able to quickly change the mode that I'm in. So I'll hit Command 7 here to switch to the Journal Mode, and now when I start SuperWhisper, it will be in Journal Mode, and I will just talk out loud, basically as if I'm talking to, you know, like a therapist. And then what the AI will output is a full transcript of everything I said verbatim, but I'll also get a therapist-style summary that basically allows me to, you know, get a glance at what I was thinking that day. It'll give me action items. I'm actually going to dive into the example here. It'll give action items that I implied or explicitly, you know, mentioned in the journal while talking. And then finally, it will ask two thought-provoking, reflective questions. And honestly, I found this like incredibly helpful. I think this is going to be a big thing that people use AI for in the future. So if you journal or want to try journaling, then definitely use this mode. Again, linked in the description. Pretty wild, right?
Like, these are all very different kinds of functionality, but it all kind of has the same sort of features at their core, and it's up to you or us as the users to create these, which I think is really fun and amazing. Now, with SuperWhisper, there are a ton of different models you can choose from, both over the internet or on your computer directly, like Whisper through OpenAI. You get the standard one for free, but the advanced ones that are going to be faster or just a little bit smarter are going to come with the pro version of SuperWhisper, which is a subscription. And that subscription is going to cost you either $8.50 per month, or if you pay by year, $85. Or you can actually get a lifetime subscription of WhisperFlow with unlimited use of cloud and local AI models for $250. I know that sounds like a lot, but AI typically costs developers a lot of money to, you know, let their users or enable their users to use. And the fact that you can get unlimited use of AI in an app for a one-time fee is quite rare, so I could actually see this going away pretty soon.
The pro version also comes with the ability to add your clipboard, record from system audio, and identify speakers. And one great part about SuperWhisper is you can change the AI for each mode. For example, if it's something relatively simple, then we can use 40 mini for my dictation prompt since that doesn't really take that much brain power. But for the meeting summarizer, we can use a more powerful model, which is going to make the output even better. MacWhisper, at least right now, does not currently support this. You basically just choose one model, which means one set of speed and intelligence for all of your prompts.
SuperWhisper also takes the personal dictionary a step further. You can add your words manually in the vocabulary section so it always gets spelling correctly, but there's also the very clever text replacement section. So instead of telling the AI exactly how to spell words, this one will let you replace them with something more useful, such as, you know, saying "Instagram link" to add your actual Instagram link, or if you're constantly sending Calendly invites, then this would be really useful for something like that.
SuperWhisper also has an iOS app, though I'm going to be honest, I didn't really test it too much because it was a little bit buggy. But basically, you should be able to create recordings and send those recordings through custom AI prompts just like you can on the computer. Now, I give the sole creator massive props for building something so powerful and so flexible. I really think this is a forward-looking tool, but the unfortunate truth is the user experience is a little bit confusing, and it was the most confusing for me to figure out how to set up. Hopefully, if you want to give this a shot, the prompts I've provided will give you a good basis, but if you want a full tutorial, then let me know. I would be happy to do that as well.
Anyway, which one's your favorite? Are you going to be using any of these, and if so, how? I would love to hear. And, um, yeah, let me know any feedback in the comments below for next time I do a deep dive on apps like these. Peace.