ninada

Stream 24,000 Hz studio speech for Classical Sanskrit and five Indian languages (sa, kn, hi, ta, te, ml) in the browser, Node.js, and Python using 27.9 MB native per-language VITS ONNX specialists (Male & Female) with zero cross-language accent bleed.

npm install ninada webml-kit

pip install ninada

Quick start

import ninada from 'ninada';
import { detectDevice } from 'webml-kit';

const { backend } = await detectDevice();
const voice = await ninada('gnumanth/ninada', {
  lang: 'kn',
  voice: 'male',
  device: backend,
});

for await (const chunk of voice.stream('ನಮಸ್ಕಾರ, ಇಂದು ಬೆಂಗಳೂರಿನಲ್ಲಿ ಹವಾಮಾನ ತುಂಬಾ ಚೆನ್ನಾಗಿದೆ.')) {
  playPcm24k(chunk.pcm, chunk.sampleRate); // Float32Array @ 24,000 Hz
}

const wav = await voice.speak('ಕನ್ನಡವೇ ಸತ್ಯ ಕನ್ನಡವೇ ನಿತ್ಯ', { lang: 'kn', voice: 'female' });
await voice.dispose();

voice.stream() yields 24,000 Hz PCM clauses as they synthesize with 0 AudioWorklet underruns and CacheStorage + IndexedDB persistence (0 refetches on reload). voice.speak() returns a complete 24 kHz RIFF WAV buffer.

Python

from ninada import Ninada

tts = Ninada(lang="sa")
for chunk in tts.stream("प्रज्ञापूर्वकव्यापारवत्त्वं जीवत्वम् । यथार्थज्ञानं प्रमाणम् ॥"):
    play(chunk["pcm"], chunk["sample_rate"])  # float32 @ 24,000 Hz (~0.197x CPU RTF)

wav_bytes = tts.speak("सत्यमेव जयते नानृतम्", lang="sa", voice="female")
open("satyam.wav", "wb").write(wav_bytes)

Ninada loads dedicated 27.9 MB W8A32+FP16 ONNX specialists (onnx/{sa,kn,hi,ta,te,ml}.onnx and onnx/{sa,kn,hi,ta,te,ml}_female.onnx), each trained exclusively on its own language's ai4bharat/Rasa single-speaker 24 kHz studio recordings and native Unicode orthography.

CLI

npx ninada "ಕನ್ನಡವೇ ಸತ್ಯ ಕನ್ನಡವೇ ನಿತ್ಯ" -l kn -o kannada.wav
ninada "सत्यमेव जयते नानृतम्" -l sa -o satyam.wav

Runs 24 kHz neural synthesis when ONNX weights are available and falls back automatically to a 0.0 MB zero-dependency Cascade Klatt-Akṣara synthesizer offline.

Native Script Orthography & Prosody

import { studioTextToIds, textToIds } from 'ninada';

const { studioIds, lang } = studioTextToIds('ವಾಙ್ಮಯ ತಪಸ್ಸು ಮತ್ತು ಜ್ಞಾನ ವಿಕಾಸ ಅತ್ಯಂತ ಮುಖ್ಯವಾದವು.', 'kn');
const { aksharas } = textToIds('धर्मक्षेत्रे कुरुक्षेत्रे समवेता युयुत्सवः ।', 'sa');

studioTextToIds() tokenizes directly in each language's native Unicode character table (sa 144 tokens, kn 85, hi 120, ta 61, te 110, ml 101) — preserving Classical Sanskrit vocalic ṛ/ṝ/ḷ, visarga sandhi, Kannada ಙ/ಃ allophony, Tamil ழ/ற/ன, and Malayalam atomic chillus (ൺ ൻ ർ ൽ ൾ ൿ) feeding a 24 kHz RelativePositionTransformer + StochasticDurationPredictor + HiFi-GAN v1 studio generator.

Benchmarks (N = 240 Held-Out 24 kHz Studio Utterances, B = 1,000 Bootstrap 95% CIs)

Per-language held-out MCD 0–8 kHz (dB, ↓): Sanskrit 3.51 (vs 5.32 MMS), Hindi 3.37 (vs 4.85 MMS), Kannada 3.35 (vs 5.04 MMS), Tamil 3.21 (vs 4.87 MMS), Telugu 3.59 (vs 5.57 MMS), Malayalam 3.34 (vs 4.88 MMS). Full bootstrap CIs and parity telemetry (0.61 dB parity MCD, r = 0.9999) in bench/SCIENCE_REPORT.md and bench/results.json.

Reproduce & Verify

PYTHONPATH=python/src:. python training/train_native_specialists_gcp.py --work-dir data/v4 --out-dir artifacts/ninada_v4
PYTHONPATH=python/src:. python bench/check_bench_thresholds.py
pytest -q && node verify-browser.mjs

Trains all 12 native 24 kHz Male/Female studio VITS specialists, evaluates 240 held-out studio clips, and verifies 0 AudioWorklet underruns and 0 cached network refetches in headless Chromium.

Links

• Live Browser Voice Studio
• Hugging Face Model (gnumanth/ninada)
• Hugging Face Dataset (gnumanth/ninada-sanskrit-indic)

License

MIT © Hemanth.HM