Tamil · Male
Murugan
Use for conversational AI, narration, customer interactions, notifications, and general speech generation.
Radiance · Tamil
Convert Tamil text into natural, expressive AI speech for voice agents, customer conversations, narration, IVR, education, media, and real-time AI applications. Use Shunya Radiance to generate spoken Tamil from native Tamil text with dedicated Tamil voices, expressive delivery, streaming and batch synthesis, multilingual speech, and production-ready audio formats.
வணக்கம். உங்கள் ஆர்டர் நாளை காலை வந்து சேரும்.
Conversational delivery · RadianceTamil TTS
Tamil speech synthesis needs to do more than pronounce characters. Natural speech depends on the complete sentence - including Tamil grammar, names, numbers, English terminology, punctuation, speaking style, and conversational context. Radiance includes dedicated Tamil voices and supports Tamil as part of Shunya's Indic speech-synthesis stack.
2
Dedicated named Tamil voices
46
Named voices across the Indic + English catalogue
11
Expression styles
<350ms
First-audio streaming latency
| Capability | Radiance · Tamil Text to Speech |
|---|---|
| Native Tamil Script | ✓ |
| Dedicated Tamil Voices | ✓ |
| Male + Female Voice | ✓ |
| Tanglish / Mixed Text | ✓ |
| Streaming Synthesis | ✓ |
| Batch Synthesis | ✓ |
| Expression Control | ✓ |
| Speed Control | ✓ |
| Voice Cloning | ✓ |
| Multiple Audio Formats | ✓ |
| API Integration | ✓ |
Try it yourself
Don't evaluate Tamil TTS only with a prepared demo sentence. Try a customer response, product name, appointment
message, notification, narration, Tamil-English sentence, or dynamically generated AI response.
Tamil voices
Radiance's named voice catalogue includes dedicated male and female speakers for Tamil.
Tamil · Male
Use for conversational AI, narration, customer interactions, notifications, and general speech generation.
Tamil · Female
Use for voice agents, customer experiences, educational material, media, and general Tamil synthesis.
Tamil language code
46 voices
One male and one female speaker across the named catalogue
| Language | Male | Female |
|---|---|---|
| English | Varun | Nisha |
| Hindi | Rajesh | Sunita |
| Tamil | Murugan | Thangam |
| Telugu | Vishnu | Lakshmi |
| Bengali | Arjun | Priyanka |
| Marathi | Siddharth | Ananya |
| Kannada | Kiran | Shreya |
| Malayalam | Krishnan | Deepa |
| Gujarati | Rakesh | Pooja |
| Punjabi | Gurpreet | Simran |
Tamil script
Tamil has its own writing system. Applications should be able to send natural Tamil text directly instead of first converting it into Roman characters.
The input can remain in native Tamil script.
Tamil script
Romanised Tamil
Ungal appointment naalai kaalai pathu manikku irukkirathu.
Both are ways of writing Tamil. Radiance speaks native Tamil script directly, so your application can send Tamil text exactly as it is written.
Tamil + English
Many Tamil speakers naturally combine Tamil and English. Especially in customer support, technology,
banking, e-commerce, workplace conversations, healthcare, and mobile applications.
உங்கள் appointment நாளை morning 10 மணிக்கு confirm ஆகிவிட்டது.
உங்கள் payment successfully process ஆகிவிட்டது.
App open ஆகுது but login button work ஆகவில்லை.
உங்கள் order dispatch ஆகிவிட்டது. Tracking link SMS-ல் வரும்.
The Shunya TTS playground supports mixed Tamil and English input.
The application does not need to manually generate separate Tamil and English audio clips for every language change.
Expressive Tamil speech
A Tamil voice agent should not necessarily sound like an audiobook narrator. A notification should
not necessarily sound like a newsreader. Radiance provides 11 expression styles.
General-purpose Tamil speech.
Natural dialogue for customer interactions and voice AI.
Storytelling and longer-form content.
Structured informational delivery.
Positive and upbeat speech.
Softer, subdued delivery.
More intense expression.
Anxious or tense delivery.
Reactive speech.
Disapproving expression.
Energetic delivery.
The sentence stays the same. How it sounds changes.
Real-time Tamil TTS
Fixed prompts can be generated ahead of time. Conversational AI needs speech dynamically.Radiance provides WebSocket streaming for low-latency speech generation.
<350ms
First-audio latency
Speech begins as synthesis progresses, so users do not wait for a long response to finish generating.
Streaming or batch
Radiance supports two synthesis modes.
Streaming
Use WebSocket synthesis when the user is waiting for the response.
Best for:
Batch
Use HTTP synthesis when the full audio file can be generated before playback.
Best for:
| Requirement | Streaming | Batch |
|---|---|---|
| Tamil Voice Agents | ✓ | - |
| Live Conversations | ✓ | - |
| Complete Audio Files | - | ✓ |
| Notifications | ✓ | ✓ |
| Narration | ✓ | ✓ |
| WebSocket | ✓ | - |
| HTTP POST | - | ✓ |
Control the audio
Radiance supports seven output formats.
| Format | Common use |
|---|---|
| PCM | Low-latency applications |
| WAV | High-quality editing |
| MP3 | Web and general delivery |
| OGG Opus | Streaming applications |
| FLAC | Lossless audio |
| μ-law | PSTN / telephony |
| A-law | SIP / telephony |
Speed control
Adjust speaking speed from 0.25× to 4.0×.
Trim silence
Trim the silence before and after generated speech, so prompts and replies start on time.
Quality
Match output quality to the channel, from compact telephony audio to high-quality files for media.
One Tamil voice - different output environments: Web · Mobile · Telephony · Media · Voice agents
Cross-lingual voices
Radiance separates voice identity from the language of the input. Shunya's documentation states that any named voice can speak any supported language. And voices associated with other languages can also synthesize Tamil.
Tamil voice
Can synthesize supported text beyond Tamil.
Tamil voice
Can synthesize supported text beyond Tamil.
Example
Voice controls the vocal character. Text + language control what is spoken. This makes it possible to maintain a chosen voice identity across multilingual experiences.
Voice cloning
Sometimes built-in voices are enough. Sometimes the voice itself is part of the product. Radiance supports voice cloning from a short reference recording.
3-6 seconds
Reference-audio range
01
Short clean recording.
02
Transcript of the sample.
03
Generate new Tamil speech using characteristics from the reference voice.
Responsible use: Only clone voices where the required rights and voice-owner permission exist for the intended use.
Tamil voice agents
A voice agent needs both a listening layer and a speaking layer.
என் order எப்போது வரும்?
Converts Tamil speech into text.
Understands the question and generates a response.
Generates spoken Tamil.
உங்கள் ஆர்டர் நாளை வந்து சேரும்.
STT → LLM → TTS
Real-world use cases
Generate dynamic Tamil speech for conversational AI.
Provide Tamil-language automated support and conversational responses.
Generate payment information, reminders, service messages, and customer communication.
Create Tamil menus, confirmations, notifications, and dynamic phone responses.
Speak appointment details, reminders, and patient-facing messages.
Speak order updates, delivery messages, return information, and customer responses.
Provide spoken Tamil interfaces for citizen-facing digital services.
Generate Tamil lessons, instructions, explanations, and educational audio.
Convert Tamil scripts and written content into spoken audio.
Provide spoken alternatives to written Tamil interfaces and information.
Developers
Radiance Indic provides HTTP batch synthesis and WebSocket streaming for Tamil. Set model to radiance-indic, language to ta, and voice to Murugan or Thangam.
{
"model": "radiance-indic",
"language": "ta",
"input": "<Conversational> வணக்கம். இன்று நான் உங்களுக்கு எப்படி உதவலாம்?",
"voice": "Murugan",
"response_format": "mp3"
}Batch endpoint
https://
Streaming endpoint
wss://
Input
From an application, content platform, customer workflow, or LLM.
Generate
Create a complete audio response through batch synthesis or receive progressive audio chunks through streaming.
Configure
Output
Built for production
A convincing Tamil audio demo is only the beginning. Production applications need dedicated voices, mixed-language handling,low-latency streaming, APIs, audio-format control, telephony support, expression, cloning, and multilingual expansion.
| Requirement | Radiance |
|---|---|
| Tamil Text to Speech | ✓ |
| Tamil Language Code | ta |
| Native Tamil Script | ✓ |
| Tamil Male Voice | Murugan |
| Tamil Female Voice | Thangam |
| Named Voice Catalogue | 46 |
| Expression Styles | 11 |
| Tanglish Input | ✓ |
| Batch Generation | ✓ |
| Streaming Generation | ✓ |
| First Audio | <350ms |
| Speed Control | 0.25×-4.0× |
| Audio Formats | 7 |
| Telephony Formats | ✓ |
| Cross-lingual Synthesis | ✓ |
| Voice Cloning | ✓ |
| API Integration | ✓ |
| Voice Agent Integration | ✓ |
Questions
Tamil Text to Speech uses speech-synthesis technology to convert written Tamil into spoken audio.
Yes. Tamil is supported by Radiance and uses the language code ta.
Shunya's current named catalogue includes Murugan (male) and Thangam (female) for Tamil.
Yes. Tamil is supported using its native Tamil script.
Yes. The Shunya TTS playground supports mixed Tamil and English input, and Radiance is built for Tanglish and other mixed-language speech.
Shunya has 46 named voices, with one male and one female speaker across its 23-language Indic catalogue plus English.
Yes. Radiance supports cross-lingual synthesis. The voice controls vocal identity while the input text controls the language being spoken.
Yes. Any voice can synthesize any supported language.
Yes. Radiance provides WebSocket streaming for real-time synthesis.
Shunya's current documentation reports first audio in under 350 ms for streaming TTS. Actual end-user latency can vary with network, infrastructure, input, and configuration.
Yes. Radiance supports 11 expression styles.
Yes. Speed can be configured from 0.25× to 4.0×.
Radiance supports PCM, WAV, MP3, OGG Opus, FLAC, μ-law, and A-law.
Yes. μ-law and A-law output formats support telephony-oriented workflows.
Radiance supports voice cloning using approximately 3-6 seconds of clean reference audio for supported voice-cloning workflows.
Yes. Streaming TTS can serve as the speech-generation layer after an LLM or conversational agent generates its Tamil response.
Yes. Radiance supports both HTTP batch synthesis and WebSocket streaming.
Get started
Don't judge Tamil TTS from our prepared sentence. Test yours. Try a customer response, Tamil name, product term, appointmentreminder, Tanglish sentence, payment message, narration, voice-agent response, or dynamically generated text.Choose Murugan or Thangam. Change the expression. Hear how Shunya turns your Tamil text into speech.