Start with a top pick.
Each voice has its own personality. These six are our most versatile starting points — hover and preview in the Studio to find your favourite.
Natural, warm — versatile across Hindi, Telugu, Marathi & Gujarati
Professional, polished — top pick for English, Kannada & Tamil
Soothing, wellness-friendly — great for Hindi & Punjabi
Neutral, versatile — the best all-round default
Authoritative — top pick for English, Tamil, Marathi & Gujarati
Natural, warm — top pick for Bengali
37 voices total — 14 female and 23 male. Every voice works across every language.
Not every voice performs equally in every language.
Use these recommendations as a starting point — they've been picked for the most natural prosody and accent match.
Two knobs. Huge range.
Pace controls how fast the voice speaks. Expressiveness controls how much prosodic variation you get — lower is steadier, higher is more human.
How fast the voice speaks. 1.0 is natural. Consistray default is 0.9.
- 0.5 – 0.7Slow, deliberateAccessibility, older listeners
- 0.8 – 0.9RelaxedEdTech, meditation, tutorials
- 1.0NaturalConversational agents, general use
- 1.1BriskNotifications, IVR, professional
- 1.2 – 1.5Fast, energeticSummaries, marketing audio
How much prosodic variation you get. Consistray default is 0.8.
- 0.1 – 0.3Flat, consistentCompliance, screen readers
- 0.4 – 0.5ControlledIVR, banking notifications
- 0.6BalancedMost use cases — a safe default
- 0.7 – 0.8Warm, expressiveCompanions, storytelling
- 0.9 – 1.0Highly expressiveEntertainment, character voices
Copy a configuration, ship faster.
Proven combinations for common scenarios. Paste the values straight into the Studio settings.
- Voice agent (Hindi)Arjun / Divya1.00.6PCM 16 kHz
- IVR / Phone systemHemant / Taniska1.10.4MULAW 8 kHz
- EdTech narrationArjun / Taniska0.90.6MP3 22 kHz
- Banking & financeShashank / Hemant1.10.3MP3 22 kHz
- Wellness & meditationPallavi / Divya0.750.5MP3 24 kHz
- News briefingHemant / Taniska1.20.5MP3 22 kHz
- StorytellingArjun / Sudha0.90.8WAV 24 kHz
- Thriller / suspenseVikram0.90.8MP3 24 kHz
How you write matters as much as which voice you pick.
Five simple habits dramatically improve output quality.
The most common mistake is romanising Indic text. Always write Indic words in their native script for the best pronunciation.
Write English words in English script and Hindi words in Devanagari — the way urban Indians actually speak.
Commas add short pauses, full stops add longer ones, and ellipses (…) create a trailing-off effect. Line breaks add natural breathing room between paragraphs.
Use commas for numbers over four digits so they read aloud correctly.
Split longer text at natural sentence boundaries before generating.
A few details that save you time.
Use it for thriller or suspense content only — it's not suitable as a general-purpose speaker.
Use the Pace and Expressiveness sliders for prosody control instead.
Keep each request under 2,500 characters for consistent quality and faster generation.