Skip to content
Other

VieNeu-TTS v3 Turbo

Fast Speech Synthesis System

2 min readupdated

VieNeu-TTS v3 Turbo is the next generation of an open-source speech synthesis and voice cloning system built from scratch for Vietnamese. The model supports instant voice cloning from a short audio clip (3–8 seconds), stable real-time streaming with time-to-first-audio starting at 115 ms, and experimental non-verbal cues and emotions like laughter or sighs. It is optimized to run on both consumer GPUs (via PyTorch with batching) and standard CPUs via ONNX Runtime. This tool is well suited for voice assistant developers and content creators producing podcasts, audiobooks, and social media videos.

Key features

  • 25 built-in voices — stable presets covering three major dialect regions with distinct vocal tones.
  • Blazing speed — RTF around 0.02 in batch mode on an RTX 3060 GPU, supporting up to 16 parallel streams.
  • Emotion tags — support for inline cues like [cười] (laughter) or [thở dài] (sigh) directly in the text.
  • OpenAI-compatible API — a ready-to-use server for easy integration via standard SDKs and tools like LiveKit.
  • LoRA support — fine-tune the model on your own voices and reading styles using included scripts.

What you can do

  • Generate studio-grade voiceovers for podcasts, lectures, or videos using the built-in voice presets.
  • Clone any voice from a short 3–8 second audio sample with automatic noise reduction.
  • Deploy your own local speech server with streaming compatible with the OpenAI SDK.
  • Fine-tune the model for a unique speaking style using the included LoRA scripts.

Example

from vieneu import Vieneu

vieneu = Vieneu()
audio = vieneu.infer("[cười] Xin chào các bạn, đây là tổng hợp giọng nói tự nhiên!", voice="Hải Đăng")
vieneu.save(audio, "output.wav")
Read moreDetails
• Sample rate: 48 kHz (high-fidelity audio)
• Performance: RTF ≈ 0.01–0.02 on GPU, RTF ≈ 0.5 on CPU
• Requirements: consumer GPU (recommended for maximum acceleration) or CPU
• Architecture: MOSS-Audio-Tokenizer-Nano codec and sea-g2p phonemizer
Video examples
Discuss Telegram 0 15

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # hoahocabc/vieneutts_pnb
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download hoahocabc/vieneutts_pnb --local-dir ~/models/vieneutts_pnb
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to vieneutts_pnb
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of vieneutts_pnb (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

hoahocabc/vieneutts_pnb
Author
hoahocabc
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

150

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments