Voice selection is more than choosing “male” or “female.” A Hindi narration voice needs to fit the audience, format and duration. A lively short-form speaker may become tiring in a twenty-minute lesson; a measured narrator may sound too formal in a conversational reply.
Microsoft’s current catalogue documents the Hindi voice IDs available through its speech service. The same IDs can be requested through the Edge online speech transport using the community edge-tts client. Google publishes its alternatives in the Cloud TTS voice list.
Audition by job
Use the exact script category you plan to publish: instructions, stories, news, dialogue or accessibility reading. Test questions, lists, abbreviations, names, amounts and pauses. Then ask native listeners to score comprehension before preference. A pleasant voice that obscures one instruction is the wrong operational choice.
Keep transport and voice separate
Edge and Azure can target the same Microsoft voice ID. Moving between them changes transport, authentication and service guarantees, not the intended character. Nkomo therefore keeps the selected voice during Edge-to-Azure fallback and reports both the provider and voice in response headers.
Make choice visible
Nkomo now returns all released Hindi voices in its capability registry. Clients can present those choices instead of hard-coding one name. The default remains a starting point; production teams can pin another available voice after native review.
The practical rule is simple: choose a voice for the content and audience, then verify that the fallback preserves it.
Comments
No comments yet.