Dictation is critical in vibe coding, and there are two components to it:
- Speech-to-text translation — Whisper is the free open model, but there’s Parakeet and other closed models too.
- Processing using an LLM model — to refine, fix typos, and format it nicely, even for
programmingVariableNames.
Both of these components can use either local or cloud models. For example, Whisper itself runs locally on your CPU/GPU.
LLM models can also run locally via Ollama, or in the cloud using your OpenAI, Claude, etc. API keys.
1. Whispr Flow
Currently the most popular app for dictation where things just work. It uses a cloud model for both speech-to-text and LLM processing/refinement.
The quality and speed are the best I’ve tried — even for vibe coding.
The annual subscription is $144, the most expensive among the three apps I’m reviewing, but I’d say it’s well worth it.
Verdict: Try it out and use my referral link to get a longer 1-month trial.
2. superwhisper
The thing about superwhisper is that it just has more glitches everywhere. Some are UI-related. Sometimes it doesn’t catch my key binding or fails to activate dictation, and then it just wasted my time..
The Mode setting isn’t very clear, but that’s where you can choose which model to use. With Pro, you can use your own API key or Ollama. Though if I’m paying for Pro, I’d just use their included paid model.
It took me a while to change the Recording Window, which is buried in Advanced Configuration. I reduced it to a mini window, but it kept showing the translated text every time. Then I finally figured out how to turn it off, though it wasn’t intuitive.
At this point, I suspect the solo founder/developer probably vibe coded the app.
A good thing is that there’s a hefty lifetime plan ($250), alongside an annual subscription ($85).
Verdict: The glitches make this a lower-quality product. Only get this if you’re on a budget. Otherwise, Whispr Flow is far superior.
3. MacWhisper
MacWhisper is another indie app, but the quality is much better than Superwhisper. I like how it positions its dictation UI — it feels smoother and more native.
It’s a one-time $74 lifetime purchase (no subscription), with the Pro version adding Parakeet and other model options.
The free tier is also generous. It lets you try multiple local speech-to-text models such as WhisperKit Large or Whisper Large v3. Or maybe not.
Interestingly, on their Pro page, the Large model is listed as a paid feature — but somehow, I’m able to use the Whisper Large model even without paying. Not sure if it’s a glitch or an extended free trial. 🤷🏻
Here’s the direct CDN download link, if you like to skip giving away your email.
A downside of MacWhisper is that it doesn’t use any LLM to post-process the text, so it won’t fix typos or format the output automatically.
Verdict: Everything feels more polished for an indie app, and the one-time $74 fee makes it pretty good value, as it also offers extra features that the others doesn’t offer — such as summarizing YouTube videos, transcribing dropped audio files, and piping audio from other apps.
The ability to use the Large Whisper model for free (TBC) makes it the best free option overall.
Conclusion
I highly recommend Whispr Flow. At least try it for free with the one-month trial. Subscribe pro if you have the budget; it’s well worth it if you do vibe coding frequently.
And also install MacWhiser. Even as a free app, their dictation and free models work much better than Apple’s!