Sanskrit text to Sanskrit speech coversion using Azure text to speech services

Hello Murthy,

Apologies for the delay in response. I understand you've been using a workaround by transliterating Sanskrit text into Telugu script to generate speech output, and you're looking for a direct Sanskrit voice to avoid pronunciation issues and streamline your workflow with Devanagari text.

After thorough verification against the official Microsoft Azure Speech service documentation, I must inform you that there is currently NO Sanskrit-specific neural voice available in Azure Text-to-Speech services. Specifically, the voice "sa-IN-MadhurNeural" does not exist in the Azure Speech portfolio.

Current Sanskrit Support Status in Azure AI Services

While Sanskrit (sa - language code) is supported in some Azure AI services, its availability varies significantly across different features:​

Sanskrit IS supported in:

• Azure AI Vision / OCR (Devanagari script recognition)​ https://docs.azure.cn/en-us/ai-services/computer-vision/language-support
• Azure AI Language (limited text processing features)​ https://learn.microsoft.com/en-us/azure/ai-services/language-service/concepts/language-support

Sanskrit is NOT supported in:

• Azure Text-to-Speech (TTS) - No neural voices available​
• Azure Speech-to-Text (STT) - No transcription models available

Recommended Workaround
1: Continue Using Hindi Voice (I would suggest this as the most practical approach)

The hi-IN-MadhurNeural voice you may have heard about can process Devanagari script text, as Hindi uses the same script. However, pronunciation will follow Hindi phonetic rules rather than authentic Sanskrit pronunciation patterns.

import azure.cognitiveservices.speech as speechsdk

# Configure Speech service
speech_config = speechsdk.SpeechConfig(
    subscription="YourSubscriptionKey", 
    region="centralindia"  # or eastus, westus, etc.
)

# Use Hindi voice for Devanagari text
speech_config.speech_synthesis_voice_name = "hi-IN-MadhurNeural"

# Synthesize Sanskrit text in Devanagari script
audio_config = speechsdk.audio.AudioOutputConfig(filename="sanskrit_output.wav")
synthesizer = speechsdk.SpeechSynthesizer(
    speech_config=speech_config, 
    audio_config=audio_config
)

# Your Sanskrit text in Devanagari
sanskrit_text = "नमस्ते भारत"

result = synthesizer.speak_text_async(sanskrit_text).get()

However, this has a few limitations:

• Pronunciation follows Hindi phonetics, not classical Sanskrit.
• Sandhi rules and Vedic accent patterns won't be accurate.
• May mispronounce Sanskrit-specific conjuncts and combinations.

Note: This is a code samples based on available documentation along with a few custom adjustments. Since environments and requirements may vary, I would kindly recommend reviewing and validating the code in a safe or test environment before applying it to production.

2: Custom Neural Voice (Enterprise Solution)

For production-grade Sanskrit TTS with authentic pronunciation, consider creating a Custom Neural Voice. https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/azure-ai-speech-text-to-speech-feb-2025-updates-new-hd-voices-and-more/4387263

This requires:​

• High-quality Sanskrit speech recordings (10-50 hours minimum)
• Professional Sanskrit voice talent.
• Azure Custom Neural Voice access (requires application approval).
• Significant investment in time and resources.

I would also suggest staying informed:

• Subscribe to Azure AI Speech updates: **Azure Speech What's New**​
• Monitor Tech Community announcements: Azure AI Speech Blog

I hope it helps a way forward.
Please accept the answer & upvote for remediation of other community members.