Transcribe
Languages & automatic detection
Find your language and understand how each transcription engine uses Automatic and your spoken-language choice.
In the appSettings → Engine → Language
On this page
Supported languages by engine
MicMyDay offers dictation in up to 99 languages. The languages available to you depend on the transcription engine and model you choose.
| Engine | Language coverage | Reference |
|---|---|---|
| Whisper | 99 languages with multilingual models. English-only models transcribe English. | Whisper large v3 model card |
| Nemotron | 80 languages. | The downloaded multilingual model’s metadata.json. |
| Parakeet v3 | 25 European languages, detected automatically. | NVIDIA model card and the full language list below. |
| Apple Speech | Varies by Mac and the language resources installed. | Apple Speech language availability. |
Cloud-provider coverage depends on the provider and model selected.
Choose the language you speak
Use Language → Spoken language with Whisper, Nemotron, Apple Speech, or a cloud provider. Choose the language of your recording, or use Automatic when your engine offers language detection. Apple Speech’s Automatic option uses your Mac’s locale.
The picker contains 29 named languages plus Automatic: Arabic, Chinese (Mandarin), Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Romanian, Russian, Slovak, Spanish, Swedish, Thai, Turkish, Ukrainian, and Vietnamese.
This picker is a set of explicit language choices, not the full language coverage of every model. A multilingual model may detect additional languages automatically. All 29 named choices work with multilingual Whisper and Nemotron models. The Norwegian choice uses Norwegian Bokmål. See language coverage by engine.
To write in a different language, use a translation rewrite profile. Spoken language describes the audio you provide.
What Automatic means
Automatic works differently depending on your engine. Selecting a named language also has a different effect.
| Engine | Automatic | A named language |
|---|---|---|
| Whisper | Detects the language from your audio. | Forces decoding in that language. |
| Nemotron | Detects among its 80 languages. | Prompts the model to decode in that language. |
| Parakeet v3 | Always detects the language from your audio. | Not applicable. Parakeet accepts no language setting. |
| Apple Speech | Uses your Mac’s current language and region. | Selects the recognizer for that language. |
| OpenAI and compatible custom endpoints | Leaves the language unspecified for the provider to determine. | Sends the language in the transcription request. |
| Gemini | Lets the model determine the spoken language. | Includes the language in the transcription settings or prompt, depending on the model. |
A spoken-language choice describes your recording. It does not translate your words. To change the output language, use a translation rewrite profile.
Whisper: choose or detect the language
Whisper’s multilingual models offer broad coverage across 99 languages. Choose Automatic to let the model identify the language, or select a language to tell it exactly how to decode your recording.
An explicit choice is useful when a short recording is being identified incorrectly. It fixes the decoding language instead of leaving Whisper to choose it. English-only downloads remain English-only regardless of this setting.
MicMyDay passes your choice directly to Whisper’s language setting. Automatic enables its built-in language detection.
Nemotron: 80 languages, with live transcription
MicMyDay downloads the multilingual Nemotron model, which covers 80 languages. Both Live and Live, steadier have the same language coverage. Changing your spoken language does not require a separate model download.
Choose Automatic to detect the language, or name it explicitly. MicMyDay matches your choice to the model’s language prompt, which directs it to decode in that language.
Live shows words sooner. Live, steadier processes slightly longer pieces of audio before showing them. This changes the timing of the live text, not the languages you can use.
Parakeet v3: supported languages
Language is detected automatically. Speak in any of the 25 languages below. Parakeet identifies it from your audio and accepts no language setting. MicMyDay replaces the Spoken language picker with an explanation while Parakeet is selected. A language saved for another engine has no effect on Parakeet.
| Language | Code |
|---|---|
| Bulgarian | bg |
| Croatian | hr |
| Czech | cs |
| Danish | da |
| Dutch | nl |
| English | en |
| Estonian | et |
| Finnish | fi |
| French | fr |
| German | de |
| Greek | el |
| Hungarian | hu |
| Italian | it |
| Latvian | lv |
| Lithuanian | lt |
| Maltese | mt |
| Polish | pl |
| Portuguese | pt |
| Romanian | ro |
| Russian | ru |
| Slovak | sk |
| Slovenian | sl |
| Spanish | es |
| Swedish | sv |
| Ukrainian | uk |
Parakeet covers 18 of the 29 named languages in MicMyDay’s shared picker. The other 7 languages in the table are available through Parakeet’s automatic detection even though they are not listed in that picker.
Arabic, Chinese, Hebrew, Hindi, Indonesian, Japanese, Korean, Norwegian, Thai, Turkish, and Vietnamese are not supported by Parakeet v3. Whisper and Nemotron support all 11. MicMyDay shows a warning if your saved language choice is one of these while Parakeet is selected.
Source: NVIDIA’s Parakeet v3 model card.
Use Apple Speech
Apple Speech is another option in the engine picker. It needs Speech Recognition permission. Automatic does not detect the language. It uses your Mac’s current language and region. If you dictate in another language, select it explicitly so Apple Speech uses the right recognizer.
Prefer on-device recognition is enabled by default. Keep it on to use local recognition for supported languages. If your language is unavailable locally, choose another engine or turn the option off to allow Apple’s server recognition where available.
Apple Speech is a transcription engine. Apple built-in under Rewrite is a separate feature for editing text.
Cloud providers and custom endpoints
Language coverage depends on the provider and transcription model. MicMyDay passes your spoken-language choice along with the recording.
For OpenAI and compatible custom endpoints, a named language is sent in the request’s language field. With Automatic, MicMyDay leaves that field out so the provider can determine the language.
Gemini receives the language and regional variant through its transcription settings for the dedicated transcription model, or through an instruction in the prompt for other models. Automatic leaves the language unspecified.
You choose a language by name in MicMyDay; the app converts it to the form each provider expects. See Providers & connections for setup.