How to Use Mallo in English on Mac
Mallo works in English out of the box. The speech models it uses — Whisper, Parakeet, and Qwen — are multilingual by design, so English just works.
Mallo works in English. No special setup needed — the speech models it uses are multilingual by design.
Mallo handles English dictation on Mac without any language configuration. Open the app, press your hotkey, and speak English.
This page explains why that is the case and what you can tune if you want to go further.
Why English just works
Mallo does not use macOS built-in dictation. It runs its own speech-to-text engine locally on your Mac, and every engine it supports — Whisper, Parakeet, and Qwen — is multilingual.
The default mode is Auto-detect. The model listens to your speech and figures out the language on its own. Speak English and it transcribes English. Speak Korean and it transcribes Korean. No switching required.
This is a meaningful difference from macOS dictation, which is tied to your system keyboard language. If your Mac is set to Korean, the built-in dictation expects Korean. Switching to English means changing your input source or keyboard layout first. Mallo skips that entirely because the model handles language detection, not the operating system.
For more on how this multilingual behavior works in practice, see Multilingual Dictation on Mac with Mallo.
If you want to optimize for English
Auto-detect works well for most people. But if you dictate almost exclusively in English, two options can tighten the workflow:
Pick an English-only model. Whisper offers Base English and Small English variants. These models are trained only on English, so they skip the language detection step entirely. That means less processing per utterance and fewer misreads on short phrases where the model might otherwise guess the wrong language.
Lock the language to English. In the same Models tab, the language dropdown lets you switch from Auto-detect to English. This tells any model — even a multilingual one like Whisper Large V3 Turbo — to expect English. It helps most when your accent or vocabulary could be interpreted as another language by Auto-detect.
Both are optional. The default setup already handles English well.
To understand the differences between engines — when Whisper is better, when Parakeet or Qwen might fit your workflow — see Choose Between Whisper, Parakeet, and Qwen.
Where English dictation works
Mallo inserts text at your cursor using voice input, so it works anywhere you can type:
- ChatGPT and Claude Code prompt boxes
- Docs and note apps
- Email and chat drafts
- Code comments and review notes
- Forms and search bars
The point is that English dictation is not limited to a specific app or a demo surface. It behaves like typing, wherever your cursor is.
Why Mallo is built this way
Mallo runs speech models locally and uses multilingual engines for the same reason: to be an input tool that fits inside your existing workflow without extra steps.
A standalone transcription app can afford to ask you to pick a language up front. An input layer that lives at your cursor cannot — any friction between thought and text defeats the purpose. That is why Mallo works as a voice input layer, and why the models it ships with handle language on their own.
If you have not read the product overview yet, What Is Mallo? is the best starting point.
FAQ
Common questions
Does Mallo support English?
Yes. All three engines — Whisper, Parakeet, and Qwen — handle English natively. Auto-detect is the default, so you can just speak English and it works.
Do I need to change my Mac system language to English?
No. Mallo uses its own speech models, not macOS dictation. Your system language does not affect how Mallo recognizes speech.
What if I switch between English and another language during the day?
Leave Auto-detect on. The model identifies the language from your speech each time, so you do not need to toggle anything between languages.
What if my English accent is not standard American or British?
The multilingual models Mallo uses — especially Whisper Large V3 Turbo — are trained on a wide range of English accents. If recognition feels off, try locking the language to English in Settings → Models to reduce ambiguity.
Related posts