R_REDDYX.XYZ
VoicePython★ 3,432⑂ 1,048 forksApache-2.0

pnnbao97/VieNeu-TTS

VIETNAMESE TTS WITH ON-DEVICE INSTANT VOICE CLONING

On-device Vietnamese TTS with instant voice cloning from 3-8s audio. Runs in real-time on CPU and GPU at 48 kHz quality.

// KEY FEATURES

  • Instant voice cloning from 3-8 sec clip
  • 25 built-in voices, conversation mode
  • Streaming 115 ms, RTF 0.01-0.02 on GPU
  • LoRA fine-tuning on consumer GPU

// DEEP DIVE: HOW IT WORKS

The project implements the v3 Turbo architecture designed from scratch: MOSS-Audio-Tokenizer-Nano codec and sea-g2p phonemizer. The model is trained on 10,000+ hours of bilingual data, outputs 48 kHz, and supports batched generation up to 32 samples including multi-speaker Conversation mode. Voice cloning works with automatic reference denoising; emotional tags like [cười] or [thở dài] are inserted directly into text.

Differentiation from alternatives — combination of on-device operation, ultra-low streaming latency (~115 ms to first audio), and OpenAI-compatible API with Docker deployment. Version v4 with higher cloning fidelity is proprietary and available only via paid API at vieneu.io; v3 Turbo remains the latest open-source release.

Requires Python and uv for installation; separate Docker profiles for CPU and GPU. Achieves ~2× real-time on CPU, on RTX 3060 up to 16 concurrent streams under 200 ms. LoRA fine-tuning possible on a single consumer GPU. Windows installer at vieneu.io, macOS coming soon.

// USE CASES

Best for: Developers building Vietnamese TTS applications

  • Add Vietnamese TTS to app via OpenAI-compatible API
  • Create multi-speaker podcast from script
  • Fine-tune user voice via LoRA on own GPU

// QUICK START

curl -LsSf https://astral.sh/uv/install.sh | sh
git clone https://github.com/pnnbao97/VieNeu-TTS.git
cd VieNeu-TTS
uv sync
# Windows: powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

Commands are taken from the repository README — check the current version on GitHub before running.

#deep-learning#on-device-ml#real-time#speech-synthesis#text-to-speech#tts#vietnamese#vietnamese-language
Open on GitHub →

New repositories every 30 minutes

REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.

Join on Telegram
← Full catalog·Full index

// SIMILAR REPOSITORIES

DOCKER WRAPPER KOKORO-82M TTS WITH OPENAI API★ 5706SMALL ENGLISH TTS MODEL 4.5M PARAMETERS★ 673LOCAL SPEECH-TO-SPEECH ASSISTANT ON GPU★ 448OFFLINE AI VOICE STUDIO IN RUST★ 312VOICE STUDIO ON MAC WITH MLX★ 247MIRLI — LOCAL AI FOR AUDIO GENERATION★ 231

← FULL CATALOG

pnnbao97/VieNeu-TTS — Vietnamese TTS With ON-Device Instant…