​
Writing Prompts in Native Scripts

Improve pronunciation, fluency, and tone of your CallHQ voice assistants by writing prompts in the native script of the target language—such as Devanagari for Hindi or accented Latin characters for French and Spanish.
CallHQ supports conversations in multiple languages out of the box. To achieve natural-sounding responses, it’s important to provide prompts in the original script of each language, instead of using transliterations written in English. Refer to our list of supported languages to check if your preferred language is available.

​
Prompting Voice Agents in Non-English Languages

When writing prompts, you can allow the agent to adjust based on the user’s language. Here’s a sample instruction for Spanish:
Maintain concise, two-sentence responses. Stick to the context at all times. Speak in Spanish by default. If the customer switches to English, respond in English from that point forward. Always mirror the customer’s language.

​
Why Use Native Scripts?

Writing prompts in native scripts ensures:
  • Better pronunciation and clarity
  • Improved language detection by the AI
  • Natural tone and structure of the output
  • Reduced risk of fallback to English pronunciation

​
Helpful Tips

  • Use tools like Google Input Tools or built-in OS keyboards for typing in different scripts.
  • Ensure accented characters are preserved (e.g. é, ñ, ç, ¿, ¡).
  • For Indian languages, prefer Unicode Devanagari or other native input formats over phonetics.

​
Examples: Native Script Prompting

❌ Poor Example:
Aapka order tayar hai, kirpya counter par le lijiye.
✅ Recommended:
आपका ऑर्डर तैयार है, कृपया काउंटर पर ले लीजिए।
Hindi voice output becomes significantly more accurate when written in the Devanagari script.

​
Common Mistakes to Avoid

Avoid phonetic spellings using English letters for prompts in other languages.
❌ Kaise ho?
✅ कैसे हो?
❌ Como estas?
✅ ¿Cómo estás?
To ensure the best experience across languages, always test your prompt using the preview feature in the CallHQ Dashboard before deploying.