# Pick the dictation language. Dimmy Help.

> Pick the dictation language. Step-by-step help for Dimmy, the native voice dictation app for macOS, Windows, and Linux.

Source: https://dimmy.app/help/language/

[← Back to Help center](https://dimmy.app/help/)

Settings Reference

# Pick the dictation language

Auto-detect or pin one of 15 languages. What the setting does depends on which engine you run: Whisper obeys it, Parakeet ignores it, Qwen3-ASR needs it.

Settings → Voice input → Language tells the speech engine what to expect. On Windows the row is labelled "Native language (you speak)". On Linux it is Settings → General → Language. The same setting also tells the filler remover which word list to use.

## Auto-detect (default)

Auto-detect ships as the default and works with local models as well as cloud ones. Dimmy passes an explicit auto value to whisper.cpp rather than leaving the language blank. Leaving it blank used to return a confident, completely empty transcript with no error at all, which read as "slow" or "stuck" rather than broken. That is fixed.

Pinning the language you actually speak is still a little quicker and a little more reliable on short clips, because the engine has nothing to work out first.

## The list

The picker offers Auto-detect plus 15 explicit languages, in two blocks. The first block is covered by both local engines. The second is marked (Whisper) because Parakeet TDT v3 does not cover it.

| Block              | Languages                                                                                         |
| ------------------ | ------------------------------------------------------------------------------------------------- |
| Both local engines | Italiano, English, Español, Français, Deutsch, Português, Nederlands, Polski, Русский, Українська |
| (Whisper) only     | 中文, 日本語, 한국어, العربية, हिन्दी                                                                     |

Auto-detect covers the rest of the 99 languages Whisper knows without listing them all in a dropdown.

## What each engine does with it

| Engine          | What the setting does                                                                      |
| --------------- | ------------------------------------------------------------------------------------------ |
| Whisper         | Sets the language token for the transcription. Auto-detect works.                          |
| Parakeet TDT v3 | Nothing at all. The engine takes no language argument, so the setting has no effect on it. |
| Qwen3-ASR       | Names the language in the instruction the model is given. Worth setting explicitly.        |
| Cloud providers | Passed on to the provider, which handles it their own way.                                 |

warning

Nothing blocks or warns you on a mismatch. Picking a (Whisper) language while Parakeet is your engine simply gives you a worse result, and the label is the only signal. On Qwen3-ASR, leaving it on Auto is the risky one: a 22-minute Italian meeting came back with its middle third in Spanish.

## Pin a language

1. Open Settings → Voice input → Language.
2. Pick a language from the dropdown.
3. It saves on its own.

## Filler removal

Filler removal has hand-tuned word lists in 6 languages: English, Italian, Spanish, French, German, Portuguese. Every other language still transcribes correctly, it just has no filler list yet. See [Filler removal](https://dimmy.app/help/filler-removal/).

## Meeting recaps detect their own language

This setting says "I speak Italian", not "this meeting was in Italian", and nobody changes it before a call with an English-speaking client. So a recap runs its own detection over the recording instead: five windows of 30 seconds, skipping the first 20 seconds, on recordings of at least a minute, and three of the five have to agree. If they do not, the recap keeps its existing wording rather than guessing.

info

If you move between languages all day, see [Multilingual workflows](https://dimmy.app/help/use-multilingual/). To work out which engine suits the languages you speak, see [Which model should you use?](https://dimmy.app/help/choose-your-model/).

Was this helpful? 

## In this category

[ Settings ReferenceSettings overview: every tab, in one mapRight-click the pill → Settings. Ten tabs grouped by purpose. Use this as the index to find anything you can configure.](https://dimmy.app/help/settings-overview/) [ Settings ReferenceAudio input: microphone and processingSettings → Voice input. Pick the input device. The processing toggles ship off, because measurement showed they made transcripts worse.](https://dimmy.app/help/audio-input/) [ Settings ReferenceStats: count, time saved, words dictatedSettings → Stats. Aggregate numbers computed from your local history. Time saved is the difference between speaking and typing the same text.](https://dimmy.app/help/stats/) [ Settings ReferenceAbout: version, updates, linksSettings → About shows the current version, runs an update check, and links to the repo, license, and acknowledgements.](https://dimmy.app/help/about-and-updates/) [ Settings ReferencePause Dimmy temporarilyDisable the hotkey without quitting the app. Useful during screen recordings, presentations, or focus blocks.](https://dimmy.app/help/pause-dimmy/)

Last updated: 2026-05-22

On this page

1. [01 Auto-detect (default)](#auto-detect-default)
2. [02 The list](#the-list)
3. [03 What each engine does with it](#what-each-engine-does-with-it)
4. [04 Pin a language](#pin-a-language)
5. [05 Filler removal](#filler-removal)
6. [06 Meeting recaps detect their own language](#meeting-recaps-detect-their-own-language)