AI voice-to-text apps have revolutionized the way we capture our thoughts, offering a more fluid and natural alternative to traditional word-for-word dictation. Unlike manual typing, generative AI apps leverage advanced language models to transcribe spoken words into coherent text, often enhancing the narrative with contextual understanding. This capability not only speeds up the writing process, but also allows for a more authentic expression of ideas.
Since I’ve upgraded to M4 Mac mini, its speed and GPU boost enable me to use AI apps better than on my Intel iMac.
In my use of AI voice-to-text apps, I’ve been SuperWhisper, Wispr Flow, Talktastic, and FixKey, allow me to do less typing and produce more notes and written content. Each app brings unique features to the table, making the choice both intriguing and challenging to decide which one I will subscribe to. Fortunately, they each have free trial or beta versions for exploring and testing.
If you haven’t yet tried voice-to-text AI apps, I strongly recommend that you do.
SuperWhisper
SuperWhisper provides access to various AI Language Models, including GPT-4o mini, Claude 3 Sonnet, Llama, and Standard v1. The voice models take your voice dictation into text output. The text outputs I receive are pretty close to what I dictate. It does not generate and add words, phrases and sentences I didn’t speak. However, it will produce grammatically correct sentences and coherent paragraphs.
SuperWhisper also allows for creating different AI prompts (instructions) that tell the AI Language model what to do with your dictation.
I created a prompt that automatically adds title to what I dictate. This unique feature is useful for when dictating in my Day One journal or Mem AI notebook.
SuperWhisper also includes a vocabulary or dictionary feature that helps the Al recognize people’s names, company names, acronyms, slang, or words from other languages. Similarly, it also has a unique Replacements feature that will output the correct spelling, acronyms, and phrases based on spoken keywords. For example, if I say “Mat blog site link,” it will output https://macautomationtips.com.
Wispr Flow
Wispr Flow (referral link) is the fastest among the apps I’ve tried, delivering smooth and quick transcriptions. However, its speed sometimes comes at the cost of clarity, with some sentence structures requiring additional editing to ensure coherence. On the other hand, voice-to-text apps require clear dictation in order to get better results. I’m finding that voice dictation takes practice to get the best results.
I mapped BetterTouchTool 4 Finger Swipe Left finger gesture to trigger Wispr Flow, so when my hand is on my trackpad, I can trigger the application. Wispr Flow typically outputs transcriptions immediately after I swipe the finger gesture to stop the recording. While I’m typing, I use the Fn or the Ctrl+Opt keys to start the recording and to get the output. (Note: notice how the in the last sentence, Wispr Flow correctly inserted “Ctrl+Opt” usage instead of spelling out the words.)
Wispr Flow also includes an Auto-Dictionary feature that auto learns words based upon your usage and adds them to its dictionary.
Wispr Flow can be used for free for up to 2000 words a week, and you can connect it to an AI language model like ChatGPT or Perplexity in order to ask questions or prompt for instructions.
By default, each of the AI apps I’m reviewing instructs you to press the Fn key to record your dictation. However, with Wispr Flow, you can create custom shortcuts for triggering the app. I use a Keyboard Maestro string trigger (jj) to activate Wispr Flow for quickly dictating one or more sentences.
I also like Flow’s user stats to see how I’m improving with dictation.
Update: Super Wisper is now available for Intel Macs. Use my referral link to check it out for free.
Talktastic
Talktastic offers a distinctive approach by expanding on dictated content, incorporating contextual elements that enhance the narrative. It provides two output formats: Ghost Writer and Transcript outputs. The Ghost Writer option transforms your verbal thoughts into polished, well-structured text, which means it will add words, phrases, sentences that you didn’t dictate but convey the context of what you were dictating. In contrast, the Transcript feature stays true to your natural speaking style, capturing your words nearly verbatim, but with grammatical corrections.
Talktastic also includes context awareness feature, so say if you are replying to an email, Talktastic takes a screenshot of what it sees in the app and then uses the context to help write the reply. Shortwave, the mail client I use, also includes a similar and powerful context awareness feature to help write and reply to emails.
Out of the four apps under review, Talktastic doesn’t yet include a dictionary feature. It does, though, allow for creating custom shortcuts to trigger recordings.
Talktastic as of this writing is still free while in beta.
FixKey
FixKey recently added a voice-to-text feature. It’s also an AI app that quickly corrects the grammar and spelling of selected text that you have manually typed (see my review here.)
I switched to using Wispr Flow over FixKey because of the better voice-to-text results. However, FixKey is still helpful for combined grammar, spelling, and voice-to-text use. While its transcriptions aren’t as accurate as other apps, its outputs are quick. You can also create a custom trigger shortcut instead of the Fn key.
Better than Apple’s Voice-to-text
Each of these AI voice-to-text apps significantly outperforms Apple’s native dictation features for Mac and iOS. The key difference is that they don’t simply transcribe spoken words – they intelligently interpret the speaker’s meaning and generate coherent, grammatically correct sentences and paragraphs. (Note: the last paragraph is Talktastic output based on what I dictated.) Super Whisper offers an iOS app, but it lacks universal integration within iOS applications. Apple’s AI should incorporate generative voice-to-text AI into its operating system, instead of simple word-for-word dictation.
Conclusion
For this article, I used Mem AI, Talktastic, and FixKey, along with ProWritingAid for final proofreading and edits. While the generative AI technology powering these applications is remarkably powerful, none of these tools can autonomously produce content without human input and oversight. I still have to do several careful edits that the apps miss, even though I’ve never been the best editor of my writing,
I am still experimenting to figure out which one of these voice detection apps I will use the most. But right now, I am leaning towards Wispr Flow and Super Wispr. It’s great that I’m able to use all four for their particular unique features.
It would be even more powerful if a dictation app could stay enabled and only produce results when it hears voice dictation. Apple’s voice-to-text apps can do that, but the feature in the Mac can take up lots of CPUs if enabled too long.
These AI voice-to-text apps have reduced my time for drafting content, writing outlines, and brainstorming ideas. Before using these, it would take twice as long to write and edit articles like this one. I now try to dictate at least 80% of what I write in my notebooks, emails, articles, online comments, and other content. The technology behind these apps will only get better with time, just as our ability to use them.
