Sri Lanka's Premier TTS Model

Rava

රාවය (Rāvaya) · "the voice, the roar, the resonance"  ·  v1.0

The most advanced Sinhala text-to-speech model ever built — carrying the language of Sri Lanka into the era of artificial intelligence.

Live synthesis activity
රාවය — ශ්‍රී ලංකාවේ හඬ, නවීන ලොවට
The Roar of Lanka — For the Modern World

Built on diffusion.
Built for Sinhala.

Rava uses a state-of-the-art diffusion-based synthesis architecture — the same class of models that power the world's leading voice AI — retrained from the ground up for the phonological complexity of the Sinhala language.

Native Sinhala Phonology
Trained on curated, high-quality Sinhala speech corpora. Our model natively handles aspirated stops, retroflex consonants, and the full phoneme inventory of standard Sinhala.
Low Latency Inference
Optimised inference pipeline delivers first audio chunk in under 200ms. Real-time streaming support for conversational applications and live broadcast scenarios.
Voice Cloning
Clone any Sinhala voice from as little as 15 seconds of reference audio. Preserves speaker identity, emotional register, and regional dialect characteristics.
Script Intelligence
Natively reads the Sinhala Unicode script. Mixed-script input — Sinhala and Latin within the same sentence — is handled without preprocessing or transliteration.
Enterprise Grade
SOC 2 compliant infrastructure. Dedicated deployment options. SLA-backed uptime. Designed for broadcasters, government services, and large-scale enterprise integrations.

Hear Sinhala,
finally heard.

Type any Sinhala text and listen to the most natural synthesis ever produced for the language. Choose from our library of certified voices.
74k
Training Hours
98%
MOS Score
<200ms
First Chunk
Rava Synthesiser

The noise is the doctrine.

Every voice begins as noise. Our visual grain honours the diffusion process — chaos becoming language, static becoming speech. Six noise fields define our aesthetic territory.

Ink Bloom
Deep Sea base · Gold corona · Primary backgrounds
Parchment Drift
Cream core · Gold edge · Documents
Temple Fire
Red Accent core · Deep Sea halo · High energy
Crimson Slash
Deep Sea base · diagonal red thrust · Bold statements
Gold Nebula
White star · Deep Sea edge · Celebration
Obsidian Calm
Deep Sea base · Faint white light · UI default

Built for Sri Lanka.
Every sector.

From state broadcasters to mobile applications — Rava is the infrastructure for Sinhala voice across every industry.

01
Broadcasting & Media
Automated Sinhala news reading, documentary narration, and real-time audio description for live television and radio.
විද්‍යුත් මාධ්‍ය
02
Education
Accessible learning tools for Sinhala-medium schools, e-learning platforms, and tools for students with visual impairments.
අධ්‍යාපනය
03
Government Services
IVR systems, public announcement automation, and accessible digital government services for all Sri Lankan citizens.
රාජ්‍ය සේවා
04
Developer API
REST API with WebSocket streaming. SDKs for Python, JavaScript, and mobile. Build the next generation of Sinhala-first applications.
සංවර්ධකයින්
The Rava Mission
The roar of Lanka,
finally heard.
රාවය — ශ්‍රී ලංකාවේ හඬ, නවීන ලොවට

This is not a brand that translates culture. It is culture — voiced with precision, warmth, and dignity, into the technologies that shape the modern world.

Integrate in
minutes.

A clean REST API with WebSocket streaming support. Your application speaks Sinhala before your first coffee is cold.

REST & WebSocket — Batch synthesis or real-time streaming. Sub-200ms first chunk latency on streaming endpoints.
Voice Library — 12 certified Sinhala voices across gender, age, and register. Custom voice cloning available on Pro tier.
SSML Support — Full SSML control over prosody, rate, pitch, and emphasis — extended with Sinhala-specific phoneme annotations.
Generous Free Tier — 10,000 characters per month. No credit card required. Built for the Sri Lankan developer community.
Python — Quick Start Copy
import rava

# Initialise the client
client = rava.Client(
    api_key="rv_your_api_key"
)

# Synthesise Sinhala text
audio = client.synthesise(
    text="ශ්‍රී ලංකාවේ හඬ",
    voice_id="nimal-standard",
    model="rava-v1",
    output_format="mp3_44100"
)

# Save or stream the result
audio.save("output.mp3")

# Or stream in real-time
with client.stream(
    text="නවීන ලොවට",
    voice_id="kamala-warm"
) as stream:
    for chunk in stream:
        play(chunk)

Aged parchment.
Temple gold. Deep ink.

A strictly defined 6-color palette drawn from Sri Lanka's visual landscape — never sterile, never corporate. Always warm. Always authoritative.

Ink
#100f0e
Cream
#f6f1ea
Red
#6f1e15
Gold
#c8a96f
Deep Sea
#1a3a5c
White
#ffffff