Shunya LabsShunya LabsPlayground

Radiance · AI Voice

AI Voice Generator

Generate natural, expressive AI voices for voice agents, customer conversations, narration, IVR, media, learning, and multilingual applications. Choose a built-in voice, control how it speaks, or create a custom voice from a short reference sample with Shunya Radiance.

46 named voices11 expression stylesVoice cloning23 + EN languages
Varun · English

Welcome to Shunya. Choose a voice, change the expression, and hear how the same sentence can sound completely different.

Conversational delivery · Radiance
Voice · VarunExpression · ConversationalEnglish · StreamingMP3 output

AI voice generation

From text to a voice that fits the Conversation

An AI voice is more than text read aloud. The speaker, pacing, expression, pronunciation, speed, language, and delivery all change how the same sentence feels. Radiance gives applications control over those layers instead of forcing every response through one fixed synthetic voice.

46

Named voices in the Radiance voice catalogue

11

Expression styles for changing voice delivery

23 + English

Named Indic and English voice catalogue

3-6 sec

Reference audio supported for voice cloning

CapabilityRadiance
AI Voice Generation✓
Male Voices✓
Female Voices✓
Expression Control✓
Voice Cloning✓
Cross-lingual Synthesis✓
Streaming Generation✓
Batch Generation✓
Speed Control✓
Multiple Audio Formats✓
Telephony Audio✓
API Integration✓

Try it yourself

Give the AI voice your Words

The best way to evaluate a synthetic voice is with the content your application will actually speak.

Choose your voice

Different experiences need different Speakers

A customer-support assistant, audiobook narrator, learning application, and notification system do not necessarily need the samevocal character. Radiance includes a catalogue of named male and female voices across English and Indic languages.

LanguageMale voiceFemale voice
EnglishVarunNisha
HindiRajeshSunita
TamilMuruganThangam
BengaliArjunPriyanka
TeluguVishnuLakshmi
MarathiSiddharthAnanya

46 named voices

The catalogue extends across 23 Indic languages and English.

Voice ≠ languageThe selected voice controls the vocal character. The input text and language configuration control what that voice speaks. That separation makes one voice useful across more than one language experience.

Expressive AI voices

One sentence. Eleven ways to Deliver it.

People do not use the same tone in every conversation. Neither should an AI voice. Radiance supports

11 expression styles for changing how speech is delivered without replacing the speaker.

Neutral

Clean, general-purpose voice delivery.

Conversational

Casual everyday speech for voice agents and customer interactions.

Narrative

Storytelling and long-form narration.

News

Formal, structured informational delivery.

Happy

Positive and upbeat speech.

Sad

Softer, more subdued delivery.

Angry

Forceful and intense speech.

Fearful

Anxious or tense expression.

Surprised

Reactive and exclamatory delivery.

Neutral

Your request has been completed successfully.

Conversational

Your request has been completed successfully.

Enthusiastic

Your request has been completed successfully.

Try the same textThe words do not change. The delivery does.

Voice cloning

Create an AI voice from a real Voice Sample

Sometimes a built-in voice is enough. Sometimes the voice itself is part of the product. Radiance supports voice cloning from a short reference recording so authorised voices can generate new sentences without recording every line manually.

3-6 seconds

Reference audio supported for voice cloning

01

Provide a reference

Upload a short, clean recording of the voice you have permission to use.

02

Create the voice

The system captures voice characteristics from the reference audio.

03

Enter new text

Type a sentence that was never present in the original recording.

04

Generate speech

Create new audio using the cloned voice.

05

Use it across languages

Where supported, the same voice identity can be used cross-lingually.

Sample voiceVoice cloneNew textNew speech

A reusable AI voice can generate content dynamically instead of depending on a fixed library of recorded sentences.

One voice, more languages

Voice identity shouldn't disappear when theLanguage changes

A multilingual assistant may speak English to one customer and Hindi to another. That should not automaticallymean becoming a completely different speaker. Radiance supports cross-lingual synthesis.The voice and the language are treated as separate controls.

Choose one voice - Varun

English

Your payment has been processed successfully.

Hindi

आपका भुगतान सफलतापूर्वक प्रोसेस हो गया है।

Tamil

உங்கள் கட்டணம் வெற்றிகரமாக செயலாக்கப்பட்டது.

Bengali

আপনার পেমেন্ট সফলভাবে প্রসেস হয়েছে।

Same voice identityDifferent supported languages. One connected voice experience. This is especially useful for multilingual assistants, customer-service platforms, and applications expanding into new markets.

Mixed-language voices

Because conversations don't always stay in one Language

Multilingual users frequently combine languages inside the same sentence. Especially in India. Radiance can generatemixed-language speech without forcing the application to divide every sentence into separate audio clips.

Hinglish

आपका payment successfully process हो गया है.

Tanglish

உங்கள் appointment tomorrow morning confirm ஆகிவிட்டது.

Benglish

আপনার order successfully confirm হয়ে গেছে.

One sentenceOne TTS workflowOne continuous voice

The result feels like one speaker having one conversation rather than multiple speech engines stitched together.

Real-time AI voices

Generate speech while the AI is Responding

Some AI voices are created in advance. Conversational AI needs to generate speech dynamically. Radiance supports both streaming and batch synthesis.

Streaming

For conversations happening now

Generate audio progressively as synthesis takes place.

Best suited for:

  • AI voice agents
  • Conversational assistants
  • Customer support
  • Real-time applications
  • Interactive experiences

Batch

For content you can generate in advance

Create the complete audio output from a complete block of text.

Best suited for:

  • Narration
  • Audiobooks
  • Learning content
  • Announcements
  • Notifications
  • Media
RequirementStreamingBatch
Voice agents✓-
Real-time interaction✓-
Complete audio files-✓
Narration✓✓
WebSocket✓-
HTTP API-✓

Control the output

The voice is only one part of the Audio

Applications also need control over how the generated speech is delivered and encoded.

Voice

Choose the vocal character.

Expression

Choose how the sentence should be spoken.

Language

Choose the language of the generated speech.

Speed

Control speaking speed from 0.25× → 4.0×.

Silence trimming

Remove unnecessary silence where tighter output is required.

Audio format

Generate audio suited to the application.

FormatCommon use
MP3Web and general applications
WAVHigh-quality audio
PCMLow-latency pipelines
OGG OpusStreaming and web
FLACLossless audio
μ-lawTelephony
A-lawTelephony

Real-world use cases

AI voices built to be Used

AI Voice Agents

Give conversational AI a natural spoken voice for live customer interactions.

Customer Support

Generate dynamic responses instead of relying entirely on pre-recorded support prompts.

IVR & Telephony

Create spoken menus, confirmations, reminders, and dynamic phone responses.

Media & Voiceovers

Generate narration for videos, articles, scripts, explainers, and digital content.

Audiobooks

Turn long-form written content into expressive spoken narration.

E-learning

Create audio versions of lessons, instructions, educational material, and learning experiences.

Notifications

Generate spoken payment alerts, order updates, appointment reminders, and account messages.

Accessibility

Provide an audio alternative to written interfaces and content.

Multilingual Applications

Use the same voice platform across different languages and markets.

AI Characters

Create consistent synthetic voices for interactive characters and digital experiences.

Voice agents

Give your AI agent a Voice

A conversational AI system has two sides. It needs to listen. And it needs to speak.

User speaks

Can you check my account balance?

Horizon

Converts speech into text.

LLM / Agent

Understands the request and generates a response.

Radiance

Generates the AI voice.

User hears

Your current available balance is…

STT → LLM → TTS

User speaksHorizonLLM / AgentRadianceUser hears

The AI voice generator becomes the speaking layer of the complete voice-agent pipeline. The response can be generated dynamically instead of selecting from a library of pre-recorded sentences.

Developers

Generate AI voices through your Application

Radiance can be integrated directly into products through APIs and streaming connections.

basic-request.json

{
  "model": "radiance-indic",
  "input": "Hello, how can I help you today?",
  "voice": "Varun",
  "language": "en",
  "response_format": "mp3"
}

Input

Text generated by

  • An LLM
  • Customer workflow
  • CMS
  • Application
  • Notification system
  • Support platform

Configure

  • Voice
  • Language
  • Expression
  • Speed
  • Output format
  • Silence trimming

Generate

Use HTTP for batch generation, or WebSocket for streaming voice generation.

Output

  • Web applications
  • Mobile apps
  • Voice agents
  • Telephony
  • Media
  • Learning platforms
  • AI products

Basic workflow

TextRadianceAI Voice
Explore the TTS API

Built-in or custom voice?

Choose based on what your application actually Needs

Choose a built-in voice when

  • You need to launch quickly.
  • You want ready-to-use male and female voice choices.
  • You do not need a unique branded speaker.
  • You want to test multiple vocal characters.
TextChoose voiceGenerate

Choose voice cloning when

  • You need a specific authorised voice identity.
  • The voice is part of your product or brand experience.
  • You want new sentences without repeatedly recording the speaker.
  • You need the same voice across changing content.
Reference voiceCreate voice identityNew textGenerate

Different workflows. Same speech platform.

Built for production

Beyond the AI voice Demo

Generating one impressive audio sample is easy to evaluate. Production voice systems have more requirements. Applications need control over voices, language, expression, speed, streaming, audio formats, APIs, deployment, and multilingual expansion.

RequirementRadiance
AI Voice Generator✓
Built-in Voices✓
Male + Female Voices✓
Expression Styles11
Voice Cloning✓
Cross-lingual Synthesis✓
Mixed-language Speech✓
Streaming Generation✓
Batch Generation✓
Speed Control✓
Multiple Audio Formats✓
Telephony Formats✓
API Integration✓
LLM → TTS✓
Self-hosted Deployment✓

Cloud or private infrastructure: Teams can use managed APIs or deploy Shunya components inside private infrastructure according to their deployment requirements. The voice generator becomes part of the application's speech infrastructure rather than a standalone demo tool.

Questions

Frequently Asked Questions

What is an AI Voice Generator?

An AI Voice Generator uses speech-synthesis technology to convert written text into spoken audio using synthetic or cloned voices.

Can Shunya generate AI voices?

Yes. Shunya Radiance generates speech using built-in voices and also supports voice cloning.

How many built-in voices does Shunya provide?

The current named Radiance catalogue includes 46 voices across 23 Indic languages and English, with male and female voice options.

Can I choose between male and female AI voices?

Yes. The named Radiance catalogue includes male and female voices for each language in that catalogue.

Can I change the emotion or style of an AI voice?

Yes. Radiance supports 11 expression styles, including Conversational, Narrative, News, Happy, Sad, Angry, Enthusiastic, and others.

Can I clone a voice?

Yes. Radiance supports voice cloning from a short reference audio sample.

How much audio do I need to clone a voice?

Radiance supports voice cloning from approximately 3-6 seconds of reference audio.

Can a cloned voice say new sentences?

Yes. After creating the voice, new text can be synthesised into speech using that voice identity.

Can one AI voice speak more than one language?

Yes. Radiance supports cross-lingual synthesis across supported languages.

Can an AI voice speak Hindi and English in the same sentence?

Radiance supports mixed-language synthesis for conversations such as Hinglish and other multilingual speech patterns.

Can I use the AI Voice Generator for voice agents?

Yes. Streaming TTS can be used as the speech-generation layer in real-time conversational applications.

Can I control speaking speed?

Yes. Radiance supports speech-speed controls from 0.25× to 4.0×.

Which audio formats can I generate?

Radiance supports PCM, WAV, MP3, OGG Opus, FLAC, μ-law, and A-law.

Can I integrate the AI Voice Generator through an API?

Yes. Radiance provides batch and streaming integration options.

Can Shunya be self-hosted?

Shunya supports self-hosted deployment for organisations that need to run the speech stack inside their own infrastructure.

Get started

Generate an AI voice with your own Words

Don't judge a voice by our prepared script. Give it yours. Try a customer response, product name, notification,narration, Hindi-English sentence, voice-agent reply, or something your application generatesdynamically. Choose a voice. Change the expression. Hear the result.