> ## Documentation Index
> Fetch the complete documentation index at: https://docs.voicetypr.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Multi-Language Support

> Use VoiceTypr with 99+ languages and automatic language detection

VoiceTypr supports transcription in **99+ languages** out of the box using Whisper AI models. No additional downloads or configuration required.

## Supported Languages

Whisper models support a comprehensive list of languages:

<AccordionGroup>
  <Accordion title="European Languages" icon="earth-europe" defaultOpen>
    English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Swedish, Norwegian, Danish, Finnish, Greek, Czech, Romanian, Hungarian, Slovak, Slovenian, Croatian, Serbian, Bulgarian, Ukrainian, Russian, Belarusian, Turkish, and more...
  </Accordion>

  <Accordion title="Asian Languages" icon="earth-asia">
    Chinese (Mandarin & Cantonese), Japanese, Korean, Hindi, Bengali, Tamil, Telugu, Thai, Vietnamese, Indonesian, Malay, Tagalog, Urdu, Punjabi, Gujarati, Kannada, Malayalam, Marathi, Nepali, Sinhala, and more...
  </Accordion>

  <Accordion title="Middle Eastern Languages" icon="mosque">
    Arabic, Hebrew, Persian (Farsi), Turkish, Urdu, Pashto
  </Accordion>

  <Accordion title="African Languages" icon="globe">
    Swahili, Hausa, Yoruba, Amharic, Somali, Shona, Afrikaans, Malagasy
  </Accordion>

  <Accordion title="Other Languages" icon="language">
    Armenian, Azerbaijani, Basque, Breton, Georgian, Icelandic, Kazakh, Latin, Mongolian, Sanskrit, Welsh, Yiddish, and more...
  </Accordion>
</AccordionGroup>

<Info>
  See the complete list of 99+ supported languages in the language selector within VoiceTypr.
</Info>

## Language Selection

### Setting Your Language

<Steps>
  <Step title="Open Models Tab">
    Navigate to the Models section in VoiceTypr
  </Step>

  <Step title="Find Language Selector">
    Located below the header, above the model list
  </Step>

  <Step title="Choose Your Language">
    Click the dropdown and select your spoken language

    Or search by typing the language name
  </Step>

  <Step title="Start Recording">
    Your language selection applies immediately to new recordings
  </Step>
</Steps>

<Tip>
  You can change language at any time. Each recording uses the language selected at the time of recording.
</Tip>

## Automatic Language Detection

Whisper can automatically detect the spoken language:

### How It Works

1. Select **"Auto"** from the language dropdown (if available on your model)
2. Start recording and speak naturally
3. Whisper analyzes the audio and detects the language
4. Transcription is returned in the detected language

<Warning>
  Auto-detection works best for:

  * Single-language recordings
  * Common languages with distinct characteristics
  * Recordings longer than 5-10 seconds
</Warning>

### When to Use Auto-Detection

<CardGroup cols={2}>
  <Card title="✅ Good Use Cases" icon="check">
    * Multilingual environments
    * Unsure which language you'll speak
    * Testing different languages
    * Quick voice notes in various languages
  </Card>

  <Card title="❌ Avoid For" icon="xmark">
    * Professional transcription
    * Maximum accuracy needed
    * Mixing multiple languages in one recording
    * Very short recordings (\< 3 seconds)
  </Card>
</CardGroup>

### Limitations

* **Code-switching**: Switching languages mid-recording may produce mixed results
* **Accents**: Strong accents may affect detection accuracy
* **Background noise**: Can interfere with language detection

<Note>
  For best results, manually select your language instead of using auto-detection.
</Note>

## Model-Specific Language Support

### Whisper Models

**All Whisper models** (Tiny, Base, Small, Medium, Large) support **99+ languages**.

**Exception:** English-only models (ending in `.en`)

* Only support English
* Faster transcription for English speech
* Higher accuracy for English
* Language selector shows only "English"

### Parakeet Models (macOS)

Parakeet models support **25 European languages** only:

<Tabs>
  <Tab title="Supported Languages">
    * Bulgarian (bg)
    * Czech (cs)
    * Danish (da)
    * German (de)
    * Greek (el)
    * **English (en)**
    * Spanish (es)
    * Estonian (et)
    * Finnish (fi)
    * French (fr)
    * Croatian (hr)
    * Hungarian (hu)
    * Italian (it)
    * Lithuanian (lt)
    * Latvian (lv)
    * Maltese (mt)
    * Dutch (nl)
    * Polish (pl)
    * Portuguese (pt)
    * Romanian (ro)
    * Russian (ru)
    * Slovak (sk)
    * Slovenian (sl)
    * Swedish (sv)
    * Ukrainian (uk)
  </Tab>

  <Tab title="Unsupported Languages">
    If you need languages outside the 25 European languages, use **Whisper models** instead.

    Whisper supports:

    * All Asian languages
    * Middle Eastern languages
    * African languages
    * And 74+ additional languages
  </Tab>
</Tabs>

<Info>
  When using Parakeet models, the language selector only shows the 25 supported European languages.
</Info>

### Soniox Cloud Models

Soniox supports **48 languages**, including:

* Major European languages
* English, Spanish, French, German, Italian, Portuguese
* Asian: Chinese, Japanese, Korean, Hindi, Thai, Vietnamese
* Middle Eastern: Arabic, Hebrew, Turkish, Persian, Urdu
* And more...

<Note>
  Language availability depends on your cloud provider. Check provider documentation for the complete list.
</Note>

## Translation to English

Whisper can translate non-English speech directly to English:

### How It Works

<Warning>
  This feature is currently not exposed in the UI but is available in the underlying Whisper model. It may be added in future versions.
</Warning>

When implemented:

1. Record in any supported language
2. Enable "Translate to English" option
3. Transcription is returned in English

**Example:**

* Speak: "Bonjour, comment allez-vous?" (French)
* Receive: "Hello, how are you?" (English)

### Use Cases

* International communication
* Learning foreign languages
* Translating voice messages
* Quick translations without switching apps

## Best Practices

<AccordionGroup>
  <Accordion title="Select Your Language" icon="language" defaultOpen>
    Always manually select your language for best accuracy:

    1. Go to Models tab
    2. Choose your spoken language
    3. Start recording

    Don't rely on auto-detection for important transcriptions.
  </Accordion>

  <Accordion title="Speak Clearly" icon="microphone">
    For non-English languages:

    * Speak at a moderate pace
    * Reduce background noise
    * Use standard pronunciation
    * Avoid heavy dialects when possible
  </Accordion>

  <Accordion title="Use Larger Models" icon="brain">
    Larger models (Medium, Large) have better multilingual performance:

    * Higher accuracy for non-English languages
    * Better handling of accents
    * More robust language detection
  </Accordion>

  <Accordion title="Test Your Language" icon="flask">
    Before important work:

    1. Test with sample recordings
    2. Verify accuracy for your accent/dialect
    3. Adjust speaking style if needed
    4. Try different models (Small vs Large)
  </Accordion>
</AccordionGroup>

## Language-Specific Tips

### Asian Languages (Chinese, Japanese, Korean)

* Use **Small model or larger** for best results
* Tiny/Base models may struggle with character accuracy
* Specify exact language (e.g., "Chinese" not "Cantonese") when available

### Tonal Languages (Mandarin, Thai, Vietnamese)

* Speak with clear tonal distinctions
* Use Medium or Large models
* Reduce background noise (tones can be subtle)

### Languages with Multiple Scripts

* Arabic: Transcribed in Arabic script
* Hindi: Transcribed in Devanagari script
* Chinese: Transcribed in Chinese characters (simplified)

<Info>
  Whisper outputs text in the native script of the language. No romanization is applied.
</Info>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Wrong Language Detected" icon="language">
    **Solution:**

    * Manually select your language (don't use auto-detect)
    * Ensure you're speaking consistently in one language
    * Try a larger model for better detection
  </Accordion>

  <Accordion title="Poor Accuracy" icon="spell-check">
    **Solution:**

    * Upgrade to a larger model (Small → Medium → Large)
    * Check microphone quality
    * Reduce background noise
    * Speak more clearly and slowly
  </Accordion>

  <Accordion title="Language Not Available" icon="ban">
    **Solution:**

    * Switch to a Whisper model (supports 99+ languages)
    * Check if using Parakeet (only 25 European languages)
    * Verify model is downloaded and selected
  </Accordion>

  <Accordion title="Mixed Languages in Output" icon="shuffle">
    **Solution:**

    * Record each language separately
    * Don't code-switch within a single recording
    * Select the primary language before recording
  </Accordion>
</AccordionGroup>

## Switching Between Languages

For users who work in multiple languages:

<Steps>
  <Step title="Before Recording">
    Open VoiceTypr settings

    Check current language selection
  </Step>

  <Step title="Change Language">
    Go to Models tab

    Select the language you'll speak
  </Step>

  <Step title="Record">
    Use your hotkey and speak in the selected language
  </Step>

  <Step title="Verify">
    Check transcription accuracy

    Adjust if needed for next time
  </Step>
</Steps>

<Tip>
  Create a workflow: Check language before important recordings. It takes 2 seconds but ensures accuracy.
</Tip>