Your Mac keeps typing the wrong words?
Before you switch apps, it's worth knowing that many dictation accuracy problems can be fixed without changing tools, and the fixes aren't always the ones people try first. Here are nine, roughly in the order we'd try them: the cheap, common causes first, the bigger changes later.
1. Fix Your Microphone First#
This is the change people skip because it isn't a software setting, and it's often the one that helps most.
A laptop's built-in microphone sits well away from your mouth and hears the whole room: echo off hard surfaces, fan noise, your own keystrokes. No model can recover words the microphone didn't capture clearly.
Better options:
- A headset or earbuds with a microphone: close to your mouth, so your voice is louder than the room. You may already own a pair. Bluetooth headsets often switch to a lower-quality audio mode while their mic is in use, so a wired headset can sound cleaner.
- A USB microphone on your desk, placed close to you rather than at arm's length.
- Whatever you choose, select it. For Apple Dictation that's System Settings → Keyboard → Dictation → Microphone source; most dictation apps have a similar setting.
Then test it: dictate the same paragraph through the built-in mic and through the new one, and compare. If the difference is obvious, you've found your main problem.
2. Add Your Vocabulary#
Speech models may never have heard of your company, your framework or your colleague Siobhán. Faced with an unfamiliar word, they substitute the nearest common one: kubectl can become "cube control" and PostgreSQL "post grey sequel".
Most dedicated dictation apps have a custom vocabulary or dictionary feature. Apple Dictation does not. If you'd rather stay with Apple's built-in tools, Voice Control has a vocabulary list (Apple's guide), but it takes over dictation while it's switched on.
Spend ten minutes adding:
- Product and company names you say often
- Technical terms, libraries, CLI tools
- Colleagues' and clients' names
- Internal acronyms
In a specialised field, this is often what turns dictation from infuriating into usable.
3. Stop Using Apple Dictation for Long-Form Work#
Apple's built-in dictation is genuinely fine for a short message. It's free, it can process speech on your Mac (Keyboard settings show which languages it handles on the device), and it needs no setup.
For a work email or a technical document, it has gaps no setting can close. There's no custom vocabulary for your terminology, nothing cleans up false starts or applies your self-corrections, and it switches itself off after 30 seconds without speech, so a long pause to think means starting again.
If most of your dictation is real work, this is where switching tools genuinely helps. See how to choose an accurate dictation app for Mac for the options.
🎙️ Tired of correcting the same words? LumeVoice has a custom dictionary and removes filler words automatically. Try it free →
4. Speak Normally, Not Slowly#
The instinct is to help the model by enunciating carefully:
"Hi… John… I… am… sorry… for… the… delay."
This usually makes accuracy worse. Speech models use the surrounding words to resolve ambiguity. Break the sentence into isolated words and you remove the context they rely on. Speak at your normal conversational pace, in complete thoughts.
5. Stop Dictating Every Comma#
Saying "comma" and "full stop" after every clause is a habit from older dictation software. Apple Dictation, with Auto-punctuation on, and AI dictation apps add the everyday marks from your phrasing and pauses.
Save spoken punctuation for the marks a tool won't guess, like a colon, quotes or brackets. Apple Dictation and LumeVoice both still understand spoken punctuation when you need it. If a tool makes you say every mark, it's worth trying one that doesn't.
6. Check Which Model You're Running#
If your app lets you choose a Whisper model (Superwhisper, MacWhisper, Buzz and VoiceInk all do), model choice matters more than most settings. Larger models are generally more accurate but slower and hungrier for memory; OpenAI's Whisper paper reports performance improving with model size, with diminishing returns for English.
| Model | Trade-off | When it's right |
|---|---|---|
| Tiny / Base | Fastest, least accurate | Quick tests, older hardware |
| Small | Middle ground | Short notes on a modest Mac |
| Medium | Slower, more accurate | Everyday dictation, if your Mac keeps up |
| Large-v3 | Most accurate, slowest, needs the most memory | Accents, jargon, noisy audio |
| Large-v3 Turbo | Much faster than Large-v3, with minimal accuracy loss (per OpenAI) | The same jobs, if Large-v3 is too slow |
Swipe to see all columns
A large model on a Mac without much memory can feel painfully slow, and running Tiny and then blaming the app for the errors is just as common. Match the model to your hardware. Cloud apps such as LumeVoice and Wispr Flow transcribe on their own servers instead, so your Mac's memory doesn't matter, but your audio does leave your Mac to be processed.
7. Fix the Room, Not Just the Mic#
Hard floors, bare walls and a large desk all create reflections that reach the microphone a moment after your voice and blur the words together.
Cheap fixes: talk closer to the mic, put something soft on the desk, and avoid dictating in an empty room with tile or wooden floors.
8. If You Have an Accent, the Engine Matters#
This is under-discussed. Speech engines are trained on different data, so one may handle your accent noticeably better than another. OpenAI's Whisper, for example, was trained on a very large, varied set of multilingual audio collected from the web, with robustness as the stated goal (the paper is titled Robust Speech Recognition via Large-Scale Weak Supervision). Still, no engine handles every accent equally well.
If you speak English as a second language and dictation has felt unusable, don't assume the problem is your pronunciation. Dictate the same paragraph into two or three tools and compare. Most offer a free tier or free words to test with; LumeVoice, for example, starts with 3,000 free words, no card required. For more on how accuracy is assessed, see AI dictation accuracy in 2026, and this guide covers dictating as a non-native speaker.
9. Decide Whether You Want Accuracy or Finished Text#
These are different problems, and mixing them up causes a lot of frustration.
A perfectly accurate transcript still contains every "um", every false start, and the sentence you abandoned halfway. That's not an accuracy failure; it's a faithful record of how people actually talk.
If what frustrates you is editing rather than errors, no amount of microphone tuning fixes it. You need a tool with a cleanup layer that removes filler words and resolves mid-sentence corrections. That's a different kind of product, not a settings change.
Quick Diagnostic#
| Symptom | Most likely cause |
|---|---|
| Random wrong words throughout | Microphone or room echo |
| Technical terms consistently wrong | No custom vocabulary |
| Dictation switches off mid-task | Apple Dictation stops after 30 seconds without speech |
| Similar-sounding words wrong ("their"/"there") | Speaking in fragments, losing context |
| Errors everywhere with your accent | Try another engine with your own voice |
| Slow, laggy local transcription | Whisper model too large for your Mac |
| Accurate but full of "um" | Working as designed: you want a cleanup layer |
Swipe to see all columns
Conclusion#
Work through it in order: microphone, custom vocabulary, then the app. It's tempting to switch tools first and stay frustrated, when the real problem was a microphone pointing at a hard wall.
If you've done all three and it still isn't right, the issue may be the last item on the list: you don't want a more accurate transcript, you want text you can send without editing.
Accuracy Is Only Half the Problem#
Getting every word right still leaves you deleting "um" and rewriting the sentence you restarted.
LumeVoice turns your speech into clean, formatted text in any app on your Mac, and on Windows, iPhone and Android too: filler words removed, self-corrections applied, and a dictionary for your terminology. Processing happens in the cloud; audio isn't stored afterwards and is never used for training. Free to start, no card needed.
Related Guides#
- How to use dictation on Mac
- How to choose an accurate dictation app for Mac
- AI dictation accuracy in 2026
- The best dictation apps for Mac
- Dictating in English as a non-native speaker
- Whisper apps for Mac: models and setup
- Android Voice Typing Not Working? 9 Fixes for Gboard and More
- Best Microphone for Dictation: Headset, USB, Lavalier or Built-In?
- Google Docs Voice Typing Not Working? 9 Fixes That Actually Help
- Windows Voice Typing Not Working? 10 Fixes in the Right Order
