DialNexa supports multiple languages per workspace. Each agent has one primary language that controls which transcription model and voice pool are used. Check the dashboard for the current language list — availability expands with provider updates.
Language reference
| Language | Code | Transcriber Support | ElevenLabs Voices | Cartesia Voices | Notes |
|---|
| English (US) | en-US | Full | Large catalog | Available | Default language |
| English (UK) | en-GB | Full | Available | Available | British accent voices |
| English (Australia) | en-AU | Full | Available | Limited | Australian accent voices |
| English (India) | en-IN | Full | Available | Limited | Indian English accent; see notes below |
| Hindi | hi | Full | Available | Limited | Devanagari script prompts recommended |
| Hindi-English (Hinglish) | hi-en | Partial | Available | Not available | Code-switching; see Writing Non-English Prompts |
| Spanish | es | Full | Available | Available | Covers Latin American and Spain variants |
| French | fr | Full | Available | Available | |
| German | de | Full | Available | Available | |
| Portuguese | pt | Full | Available | Limited | Brazilian and European variants |
| Arabic | ar | Partial | Available | Not available | MSA; dialect support varies by voice |
| Mandarin Chinese | zh | Partial | Available | Not available | Simplified characters; tonal accuracy varies |
Language support depends on both the transcription model and the selected voice. A language marked “Full” for transcription means Deepgram Nova-2 has strong accuracy data for that language. “Partial” means the language is supported but has higher word error rates or requires specific model configuration.
Language-specific notes
Indian English (en-IN)
Indian English has distinct phonological patterns that affect transcription accuracy with models trained primarily on North American English. DialNexa recommends:
- Use the Nova-2 (Phone Call) model, which has better generalization to non-North American accents
- Set Denoising Mode to Low rather than High — aggressive denoising can attenuate certain Indian English phonemes
- Test with real callers from your target region before going live
Hindi (hi)
- Deepgram supports Hindi transcription. Accuracy is higher for formal (Shuddha) Hindi than for heavily code-switched speech.
- For mixed Hindi-English callers, use the Hinglish (
hi-en) language setting instead.
- ElevenLabs provides Hindi-capable voices; check the voice selector with language filter set to Hindi for available options.
Hindi-English / Hinglish (hi-en)
Hinglish is not a separate language but a code-switching pattern where speakers alternate between Hindi and English within a sentence. DialNexa handles this with a specialized transcriber configuration and the Hinglish Map in Speech Settings.
- Transcriber support is partial: the system handles common switching patterns but may struggle with rapid or idiosyncratic switching.
- Voice synthesis for Hinglish uses English-capable voices with Hindi pronunciation rules applied where needed.
- See Writing Non-English Prompts for prompt strategy.
Arabic (ar)
- DialNexa’s Arabic support targets Modern Standard Arabic (MSA). Dialectal Arabic (Egyptian, Gulf, Levantine, Moroccan) has lower accuracy.
- Cartesia does not support Arabic. Use ElevenLabs with an Arabic-capable voice.
- For dialectal Arabic use cases, contact support to discuss custom transcription configurations.
Mandarin Chinese (zh)
- Transcription is provided via Deepgram and targets Simplified Chinese. Traditional Chinese characters may be transcribed but accuracy is lower.
- Tonal accuracy in TTS depends heavily on the selected ElevenLabs voice — preview voices carefully before deployment.
- Cartesia does not support Mandarin.
Voice provider language coverage
| Provider | Best language coverage | Limited coverage |
|---|
| ElevenLabs | EN (all variants), ES, FR, DE, PT, HI, AR | ZH, and other languages via Multilingual v2 model |
| Cartesia | EN-US, EN-GB, ES, FR, DE, PT | EN-IN, HI, AR, ZH not supported |
Checking the current language list
The dashboard language list is updated as new languages and models become available. To see what is currently available in your workspace:
- Open any agent’s settings
- Go to Speech > Language
- The dropdown shows all languages currently available for selection in your workspace
