Skip to main content

Supported Languages

Kallglot supports real-time transcription and translation for 25+ languages. This page lists all supported languages and their capabilities.

Language Codes

We use ISO 639-1 language codes. For regional variants, we use ISO 639-1 combined with country codes (e.g., de-CH for Swiss German).

Fully Supported Languages

These languages support all Kallglot features: transcription, translation, and text-to-speech.
LanguageCodeTranscriptionTranslationTTS Voices
Englishen6
Germande4
Frenchfr4
Spanishes4
Italianit3
Portuguesept3
Dutchnl2
Polishpl2
Russianru2
Japaneseja3
Koreanko2
Chinese (Mandarin)zh3
Arabicar2
Hindihi2
Turkishtr2

Regional Variants

For more accurate transcription, you can specify regional variants:
VariantCodeBase Language
Swiss Germande-CHGerman
Austrian Germande-ATGerman
Swiss Frenchfr-CHFrench
Canadian Frenchfr-CAFrench
Swiss Italianit-CHItalian
Brazilian Portuguesept-BRPortuguese
European Portuguesept-PTPortuguese
Latin American Spanishes-LASpanish
Castilian Spanishes-ESSpanish
British Englishen-GBEnglish
American Englishen-USEnglish
Australian Englishen-AUEnglish

Using Regional Variants

{
  "mode": "bidirectional_translation",
  "source_language": "de-CH",
  "target_language": "fr-CH"
}

Transcription-Only Languages

These languages support transcription but not text-to-speech synthesis:
LanguageCodeTranscriptionTranslation
Greekel
Czechcs
Romanianro
Hungarianhu
Swedishsv
Norwegianno
Danishda
Finnishfi
Thaith
Vietnamesevi
Indonesianid
Malayms
Hebrewhe
Ukrainianuk
For transcription-only languages, translated text is available but spoken audio output uses a fallback voice in the same language family.

Language Detection

Kallglot can automatically detect the spoken language when you don’t know it in advance:
{
  "mode": "bidirectional_translation",
  "source_language": "auto",
  "target_language": "en"
}

Auto-Detection Limitations

  • Initial detection may take 2-3 seconds
  • Works best when speaker uses a single language
  • Accuracy improves with more audio

Detected Language Event

When using auto-detection, you’ll receive a language_detected event via WebSocket:
{
  "type": "language_detected",
  "speaker": "customer",
  "language": "de",
  "confidence": 0.94
}

Translation Pairs

Not all language pairs are equally supported. Some pairs have higher quality due to more training data.

Tier 1 Pairs (Highest Quality)

  • English ↔ German
  • English ↔ French
  • English ↔ Spanish
  • English ↔ Italian
  • English ↔ Portuguese
  • English ↔ Dutch
  • English ↔ Japanese
  • English ↔ Chinese
  • German ↔ French
  • Spanish ↔ Portuguese

Tier 2 Pairs (High Quality)

All combinations between:
  • English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian

Tier 3 Pairs (Good Quality)

All other supported language combinations use English as a pivot language for translation.

Voice Options by Language

English Voices

Voice IDNameStyle
en-alloyAlloyNeutral, professional
en-echoEchoWarm, friendly
en-fableFableExpressive
en-onyxOnyxDeep, authoritative
en-novaNovaEnergetic
en-shimmerShimmerCalm, soothing

German Voices

Voice IDNameStyle
de-annaAnnaProfessional
de-markusMarkusFriendly
de-sarahSarahWarm
de-thomasThomasAuthoritative

Spanish Voices

Voice IDNameStyle
es-elenaElenaProfessional
es-miguelMiguelFriendly
es-luciaLuciaWarm
es-carlosCarlosAuthoritative
Use the voice parameter when creating AI agent sessions to select a specific voice.

Best Practices

When you know the language, always specify it rather than using auto-detection. This improves accuracy and reduces latency.
For Swiss customers, use de-CH, fr-CH, or it-CH for better recognition of regional accents and vocabulary.
Test your specific language pair before going live. Some pairs may need fine-tuning for domain-specific terminology.
For specialized vocabulary (medical, legal, technical), create custom glossaries to ensure consistent translation.

Requesting New Languages

Need a language that’s not listed? Contact us at support@kallglot.com to request new language support.