Multi-Domain Neural Machine Translation, VITS Speech & Lexicon Engine

Rigwe Language AI Showcase

High-performance inference engine for English to Rigwe translation, speech-to-text voice recognition, natural VITS speech synthesis, and interactive dictionary lookup.

NMT Latency

38.5ms

CTranslate2 INT8 / NLLB-200

TTS Synthesis

112ms

VITS + Rigwe G2P Engine

Pipeline E2E

150.5ms

Single-Pass Inference

Memory Footprint

1.4GB

Quantized Model Execution

Presets:

English (Input)64 chars

Rigwe (Translation Output)

Click Translate & Synthesize to generate real-time Rigwe text and speech waveform.

Rigwe Speech Audio Output22.05 kHz WAV

0.0s / 0.0s

CTranslate2 NMT Engine

Accelerated FP16/INT8 quantized inference execution yielding sub-40ms translation latency on standard CPU instances.

VITS TTS + G2P Rules

Phonemized end-to-end VITS acoustic generator converting orthographic text into 22.05kHz speech waveforms.

Vector Search Corpus

SentenceTransformers semantic vector retrieval indexing multi-domain parallel text across greetings, education, and culture.