🦜 VieNeu-TTS

VieNeu-TTS is an advanced on-device Vietnamese Text-to-Speech (TTS) with instant voice cloning and English–Vietnamese bilingual support. The SDK defaults to VieNeu-TTS v3 Turbo (48 kHz) and the minimal install is torch-free — on CPU it runs entirely on ONNX Runtime.

Hugging Face v3 Turbo License

✨ Key Features

  • v3 Turbo, 48 kHz — high-fidelity, natural Vietnamese speech (default).
  • Torch-free on CPU — minimal install runs on ONNX Runtime; PyTorch is never imported.
  • fp32 backbone by default on CPU — maximum fidelity. Use Vieneu(precision="int8") for ~1.6× speed & ~4× smaller download (needs a VNNI-capable CPU).
  • Built-in default voices — call them by name, no reference clip needed.
  • Instant voice cloning — clone any voice from 3–5s of audio.
  • Emotion cues (experimental) — drop [cười], [thở dài], [hắng giọng] into the text.
  • Bilingual (En–Vi) code-switching, fully offline.

📦 Install

CPU (default) — torch-free, runs v3 Turbo via ONNX Runtime. Most users want this:

pip install vieneu

GPU (CUDA) — only if you have an NVIDIA GPU. Install a CUDA build of PyTorch yourself first. Batching then turns on automatically on CUDA — same API, no code change:

pip install torch==2.8.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu128
pip install "transformers==4.57.6"   # Qwen3 backbone + MOSS codec (pinned — most stable for the GPU SDK)
pip install vieneu

ℹ️ When is GPU actually worth it? The GPU win comes from batching, so it only pays off on long text (many chunks generated together in one forward — long-form or bulk synthesis). For short text the torch-free CPU/ONNX path is usually faster (there's no batch to fill). Use CPU for short, interactive calls; reach for GPU for long-form or high-throughput work.

🚀 Quick Start (Python SDK)

from vieneu import Vieneu

# Default = v3 Turbo (48 kHz). GPU → PyTorch (auto-detected).
# On CPU the backbone runs fp32 by default (max quality); pass precision="int8" for speed (VNNI CPU).
vieneu = Vieneu()                    # fp32 backbone (default, max quality)
# vieneu = Vieneu(precision="int8")  # int8 backbone (faster on CPU with VNNI, ~4x smaller)
# 💡 On a GPU machine you can still switch to ONNX/CPU if you prefer: Vieneu(backend="onnx")

# 1. Built-in voice by name — no reference needed
print("🔊 Generating speech...")
audio = vieneu.infer("Xin chào, đây là VieNeu-TTS.", voice="Adam")
vieneu.save(audio, "output.wav")
print("✅ Saved to output.wav")

# List the built-in voices
voices = vieneu.list_preset_voices()
print(f"\n🎙️  {len(voices)} built-in voices available:")
for label, voice_id in voices:
    print(f"  - {label} ({voice_id})")

# 2. Reading style: DEPRECATED on v3 Turbo — `style` is accepted but IGNORED. The style
#    is already implied by the reference (preset voice / cloned clip), so output is
#    always the natural reading style. Just pick the voice:
audio = vieneu.infer("Bản tin sáng nay.", voice="Adam")

# 3. Emotion / non-verbal cues — EXPERIMENTAL: [cười] [thở dài] [hắng giọng]
audio = vieneu.infer("Nghe hay quá đi [cười].", voice="Adam")

# 4. ⚡ Batch on GPU: infer_batch() runs many texts in ONE batched forward — same API.
#    On a CUDA GPU the chunks from every text share each forward step (big throughput
#    win); on CPU it still works (no error), just sequentially. Batch caps at
#    max_batch_size (default 32; or infer_batch(..., batch_size=64); batch_size=1
#    disables). A single long infer() also auto-batches its own chunks. Uncomment to try:
#
# import time
# texts = [
#     "Chào cả nhà, hôm nay mình sẽ hướng dẫn các bạn cách cài đặt và sử dụng bộ giọng đọc mới.",
#     "Giọng nghe cực kỳ tự nhiên và truyền cảm, lại có thể chuyển đổi biểu cảm một cách linh hoạt.",
#     "Nếu thấy hữu ích, các bạn nhớ để lại một lượt thích và chia sẻ video này cho mọi người nhé!",
# ] * 10   # 30 texts — enough to fill the batch and really show the GPU throughput win
# t0 = time.time()
# audios = vieneu.infer_batch(texts, voice="Adam")
# elapsed = time.time() - t0
# total_audio = sum(len(a) for a in audios) / 48_000
# print(f"⚡ {len(texts)} texts | audio {total_audio:.1f}s | wall {elapsed:.1f}s | RTF {elapsed/total_audio:.3f}")
# for i, a in enumerate(audios):
#     vieneu.save(a, f"batch_{i}.wav")

🔊 Real-time streaming

v3 Turbo streams frame-by-frame (first audio ~300 ms, RTF < 1 on CPU). Streaming runs on the ONNX/CPU engine — the GPU/PyTorch engine is for batch throughput, not streaming, so pin backend="onnx" for realtime. Iterate infer_stream:

vieneu = Vieneu(backend="onnx")   # force ONNX/CPU — the streaming path (int8)
for chunk in vieneu.infer_stream("Xin chào các bạn!", voice="Adam"):
    play(chunk)   # np.float32 @ 48 kHz, play/write as it arrives

A full FastAPI web demo is in apps/web_stream.py (uv run python -m apps.web_stream → http://localhost:8001).

🦜 Zero-shot Voice Cloning

Clone from a short clip; the reference is auto-denoised and trimmed to ≤ 8s.

from vieneu import Vieneu
vieneu = Vieneu()

# Clone straight from a 3–8s clip
audio = vieneu.infer("Chào bạn, đây là giọng của tôi.", ref_audio="path/to/voice.wav", denoise=True)
vieneu.save(audio, "cloned.wav")

# Save a cloned voice and reuse it by name
vieneu.add_voice("Giọng của tôi", "path/to/voice.wav")
audio = vieneu.infer("Câu này dùng giọng đã lưu.", voice="Giọng của tôi")

# Just clean up a clip (no synthesis)
wav, sr = vieneu.denoise("noisy.wav", out_path="clean.wav")

denoise, add_voice, and cloning work on every backend, including the torch-free CPU/ONNX install.

🔬 Model Overview

Model Engine Device Sample Rate Features
VieNeu-TTS v3 Turbo (default) ONNX (CPU) / PyTorch (GPU) CPU/GPU 48 kHz Default voices, cloning, emotion cues

🤝 Support & Links

Made with ❤️ for the Vietnamese TTS community

Project links

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (3.6.3)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for pnnbao from gravatar.com pnnbao

Credits

Author: Phạm Nguyễn Ngọc Bảo

License

Apache Software License (Apache License)

Requires

Python >=3.10

Provides Extra

legacy pdf watermark

Tags

text-to-speech tts vietnamese voice-cloning speech-synthesis real-time on-device

Classifiers

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

vieneu-3.4.0.tar.gz (1.2 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

vieneu-3.4.0-py3-none-any.whl (1.2 MB view details)

Uploaded Python 3

File details

Details for the file vieneu-3.4.0.tar.gz.

File metadata

  • Download URL: vieneu-3.4.0.tar.gz
  • Upload date:
  • Size: 1.2 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for vieneu-3.4.0.tar.gz
Algorithm Hash digest
SHA256 92132511d5997e2bee2b422662491a4d22416f5a20d23f97cb68f3bcd6508bea
MD5 14a0249acd6af06127b24c73f8d0585f
BLAKE2b-256 e78ef62e69932d94c4326a242cf041f3e4188145f48ca1a0c680e14494718163

See more details on using hashes here.

File details

Details for the file vieneu-3.4.0-py3-none-any.whl.

File metadata

  • Download URL: vieneu-3.4.0-py3-none-any.whl
  • Upload date:
  • Size: 1.2 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for vieneu-3.4.0-py3-none-any.whl
Algorithm Hash digest
SHA256 0a308ff363839964743b344499e722cea7e9a75c45c34cfd4dea77d25e533690
MD5 910fc6dae8931a90c3ac2484d1b550e4
BLAKE2b-256 317a698d23881f59bd5f65cbc813b631f0bb0580dcfcea3bb9ea33d6ae024e95

See more details on using hashes here.

Release history Release notifications | RSS feed

3.6.3

2 files

3.6.2

2 files

3.6.1

2 files

3.6.0

2 files

3.5.4

2 files

3.5.3

2 files

3.5.2

2 files

3.5.1

2 files

3.5.0

2 files

This release

3.4.0 This release

2 files

3.3.0

2 files

3.2.12

2 files

3.2.11

2 files

3.2.10

2 files

3.2.9

2 files

3.2.8

2 files

3.2.7

2 files

3.2.6

2 files

3.2.5

2 files

3.2.4

2 files

3.2.3

2 files

3.2.2

2 files

3.2.1

2 files

3.2.0

2 files

3.1.4

2 files

3.1.3

2 files

3.1.2

2 files

3.1.1

2 files

3.1.0

2 files

3.0.13

2 files

3.0.12

2 files

3.0.11

2 files

3.0.10

2 files

3.0.9

2 files

3.0.7

2 files

3.0.6

2 files

3.0.5

2 files

3.0.4

2 files

3.0.3

2 files

3.0.2

2 files

3.0.1

2 files

3.0.0

2 files

2.7.0

2 files

2.6.1

2 files

2.6.0

2 files

2.5.0

2 files

2.4.3

2 files

2.4.2

2 files

2.4.1

2 files

2.4.0

2 files

2.3.0

2 files

2.2.0

2 files

2.1.3

2 files

2.1.2

2 files

2.1.1

2 files

2.1.0

2 files

2.0.2

2 files

2.0.1

2 files

2.0.0

2 files

1.3.0

2 files

1.2.9

2 files

1.2.8

2 files

1.2.7

2 files

1.2.6

2 files

1.2.5

2 files

1.2.4

2 files

1.2.3

2 files

1.2.2

2 files

1.2.1

2 files

1.2.0

2 files

1.1.9

2 files

1.1.8

2 files

1.1.7

2 files

1.1.6

2 files

1.1.5

2 files

1.1.4

2 files

1.1.3

2 files

1.1.2

2 files

1.1.1

2 files

1.1.0

2 files

1.0.7

2 files

1.0.6

2 files

1.0.4

2 files

1.0.3

2 files

1.0.2

2 files

1.0.1

2 files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page