SAVRN
Search Contact SAVRN

Open-weight model · Text to speech

VieNeu-TTS-v3-Turbo

by Pham Nguyen Ngoc Bao pnnbao-ump/VieNeu-TTS-v3-Turbo

VieNeu-TTS v3 Turbo is the next generation of Vietnamese TTS — 48 kHz high-fidelity speech, 23 built-in preset voices across three regions (North / Central / South), instant voice cloning, real-time streaming with an OpenAI-compatible API (16 concurrent…

Parameters131M
Context1,024
Weights743.2 MB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads587.9k

Runs On

What it takes to serve VieNeu-TTS-v3-Turbo (131M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.3 GB 0.3 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.1 GB 0.2 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.1 GB 0.1 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on VieNeu-TTS-v3-Turbo

Put this one on a CPU box first. Its vieneu SDK, v3.7.1, runs on ONNX Runtime without PyTorch on CPU and switches to a PyTorch engine on CUDA. Pham Nguyen Ngoc Bao's Vietnamese text-to-speech model carries 23 preset voices across the North, Central and South regions, takes inline emotion cues and moves between English and Vietnamese mid-sentence, all at 48 kHz. The 16-bit weights are 0.3 GB and need 0.3 GB of memory, and the publisher cites 16 concurrent streams on one RTX 3060, so the cheapest Index slot, one MI300X at $1.85 an hour, is more card than one service needs.

Apache 2.0 allows commercial use, modification and redistribution, with an express patent grant. Context is 1,024 tokens, so long scripts get chunked upstream. Training data is the publisher's VieNeu-TTS-10k-ENVI set; read what it holds before cloning customer voices, and pin a version, since the last update landed on 2026-09-16.

SAVRN Research, 2026-09-18

Model Card

By Pham Nguyen Ngoc Bao, published under apache-2.0, revision 5f2a3e93092e.

Overview

VieNeu-TTS v3 Turbo is the next generation of Vietnamese TTS — 48 kHz high-fidelity speech, 23 built-in preset voices across three regions (North / Central / South), instant voice cloning, real-time streaming with an OpenAI-compatible API (16 concurrent streams on one RTX 3060), inline emotion cues, and seamless bilingual (En–Vi) code-switching.

The reference implementation is the vieneu Python SDK (v3.7.1). Its minimal install is torch-free: on CPU everything runs on ONNX Runtime (PyTorch is never imported), and on a CUDA machine it auto-switches to the PyTorch engine with automatic batching and a continuous-batching stream scheduler — same API, no code change.

Read the full model card (1,902 words)

Configuration

Architecture
VieNeuV3TurboForTTS
Context length (tokens)
1,024
Layers
12
Hidden size
768
Feed-forward size
3,072
Attention heads
12
Key/value heads
4
Head dimension
64
RoPE base
10000
Model type
vieneu_v3_turbo

Identity and Version

Repository
pnnbao-ump/VieNeu-TTS-v3-Turbo
Publisher
Pham Nguyen Ngoc Bao
Task
Text to speech
Modality
Audio
Library
Not stated by the source
Parameters
131M parameters
Languages
vi, en
Revision
5f2a3e93092efaba9153253ff5f2e6a8e810e4f2
First published
2026-06-05
Last updated
2026-09-16

Files and Weights

33 files, 1.7 GB in total. The weights are 16 files totalling 743.2 MB in npz, onnx, safetensors.

Weights16 files · 743.2 MB
Configuration6 files · 10.0 KB
Tokenizer6 files · 106.4 KB
Documentation1 file · 15.5 KB
Other3 files · 934.5 MB
Repository1 file · 1.7 KB
Every file
FileTypeSizeSHA-256
denoiser.onnxWeights42.7 MB b7621953291c
model.safetensorsWeights261.8 MB 0ea96dbb5a76
onnx/vieneu_acoustic_cached.onnxWeights14.3 MB df1995fea383
onnx/vieneu_decode_step.onnxWeights306.1 KB bedc379cea61
onnx/vieneu_prefill.onnxWeights324.5 KB 27f8b064f6b5
onnx/vieneu_v3_heads.npzWeights25.8 MB 37c416aecea3
onnx_int8/vieneu_acoustic_cached.onnxWeights7.2 MB f631e3387c78
onnx_int8/vieneu_decode_step.onnxWeights1.1 MB 2c5b30bd8ccb
onnx_int8/vieneu_prefill.onnxWeights1.1 MB c6a80dabf67c
onnx_int8/vieneu_v3_heads.npzWeights52.2 MB fb22484baa42
onnx_update/vieneu_acoustic_cached.onnxWeights7.2 MB f631e3387c78
onnx_update/vieneu_decode_step.onnxWeights306.1 KB bedc379cea61
onnx_update/vieneu_prefill.onnxWeights324.5 KB 27f8b064f6b5
onnx_update/vieneu_v3_heads.npzWeights52.2 MB fb22484baa42
speaker_encoder.onnxWeights28.3 MB a6ac6a639977
update/model.safetensorsWeights248.0 MB 119003a9e121
config.jsonConfiguration1.6 KB —
onnx_int8/config.jsonConfiguration2.2 KB —
onnx_update/config.jsonConfiguration2.2 KB —
special_tokens_map.jsonConfiguration855 B —
update/config.jsonConfiguration2.2 KB —
update/special_tokens_map.jsonConfiguration1.1 KB —
README.mdDocumentation15.5 KB —
onnx/vieneu_backbone_shared.dataOther415.3 MB 6f28d66062bb
onnx_int8/vieneu_backbone_shared.dataOther103.9 MB bb683925f7c8
onnx_update/vieneu_backbone_shared.dataOther415.3 MB c7c072193db3
.gitattributesRepository1.7 KB —
onnx_int8/tokenizer.jsonTokenizer22.3 KB —
onnx_update/tokenizer.jsonTokenizer22.3 KB —
tokenizer.jsonTokenizer22.4 KB —
tokenizer_config.jsonTokenizer8.4 KB —
update/tokenizer.jsonTokenizer22.3 KB —
update/tokenizer_config.jsonTokenizer8.6 KB —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
743.2 MB
Download from Pham Nguyen Ngoc Bao

Released by Pham Nguyen Ngoc Bao through its official repository on Hugging Face. Read the license.

Built From

  • Trained on (disclosed) pnnbao-ump/VieNeu-TTS-10k-ENVI

Memory Requirements

PrecisionWeights in memory
As published743.2 MB
16-bit0.3 GB
8-bit0.1 GB
4-bit0.1 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About VieNeu-TTS-v3-Turbo

How much GPU memory does VieNeu-TTS-v3-Turbo need?

About 0.3 GB at 16-bit and 0.1 GB at 4-bit: the weights (131M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run VieNeu-TTS-v3-Turbo on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use VieNeu-TTS-v3-Turbo commercially?

Yes. VieNeu-TTS-v3-Turbo is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

What is VieNeu-TTS-v3-Turbo's context length?

1,024 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Text to speech

Soprano-1.1-80M

Eugene Kwek

2026.01.14 - Soprano-1.1-80M released! 95% fewer hallucinations and a 63% preference rate over Soprano-80M. 2026.01.13 - Soprano-Factory released! You can now train/fine-tune your own Soprano models. 2025.12.22 - Soprano-80M released! Code | Demo Soprano is an ultra‑lightweight, on-device text‑to‑speech (TTS) model designed for expressive, high‑fidelity speech synthesis at unprecedented speed. Soprano was designed with the following features: - Up to 2000x real-time generation on GPU and 20x real-time on CPU - Lossless streaming with one plus one, etc) If Soprano produces unsatisfactory results, you can easily regenerate it for a new, potentially better generation. You may also change the…

Open weights apache-2.0 80M parameters 1,024 tokens transformers
View model

Model · Text to speech

mms-tts-hin

AI at Meta

mms - vits pipelinetag: text-to-speech This repository contains the Hindi (hin) language text-to-speech (TTS) model checkpoint. This model is part of Facebook's Massively Multilingual Speech project, aiming to provide speech technology across a diverse range of languages. You can find more details about the supported languages and their ISO 639-3 codes in the MMS Language Coverage Overview, and see all MMS-TTS checkpoints on the Hugging Face Hub: facebook/mms-tts. MMS-TTS is available in the Transformers library from version 4.33 onwards. VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is an end-to-end speech synthesis model that predicts a speech…

Open weights cc-by-nc-4.0 36M parameters transformers
View model

Model · Text to speech

mms-tts-eng

AI at Meta

mms - vits pipelinetag: text-to-speech This repository contains the English (eng) language text-to-speech (TTS) model checkpoint. This model is part of Facebook's Massively Multilingual Speech project, aiming to provide speech technology across a diverse range of languages. You can find more details about the supported languages and their ISO 639-3 codes in the MMS Language Coverage Overview, and see all MMS-TTS checkpoints on the Hugging Face Hub: facebook/mms-tts. MMS-TTS is available in the Transformers library from version 4.33 onwards. VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is an end-to-end speech synthesis model that predicts a speech…

Open weights cc-by-nc-4.0 36M parameters transformers
View model

Model · Text to speech

mms-tts-bam

AI at Meta

mms - vits pipelinetag: text-to-speech This repository contains the Bamanankan (bam) language text-to-speech (TTS) model checkpoint. This model is part of Facebook's Massively Multilingual Speech project, aiming to provide speech technology across a diverse range of languages. You can find more details about the supported languages and their ISO 639-3 codes in the MMS Language Coverage Overview, and see all MMS-TTS checkpoints on the Hugging Face Hub: facebook/mms-tts. MMS-TTS is available in the Transformers library from version 4.33 onwards. VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is an end-to-end speech synthesis model that predicts a speech…

Open weights cc-by-nc-4.0 36M parameters transformers
View model

Model · Text to speech

kokoro-inno-clone-tuner

Jeremy Braun

Zero-shot voice tuner for Kokoro-82M. Outputs base Kokoro compatible voice packs @ [510, 1, 256]. Same passage for every voice, enrolled from the references. LibriTTS-R speakers are dev-clean held out from training. Integrated into Kokoro-FastAPI (v0.9.0+) The pack is a plain tensor; torch.save(pack, "voices/amme.pt") makes it a voice file like any other, prefixed by accent and gender like the stock packs. The pitch-tracking ceiling is set automatically from the reference's harmonic spacing, so band-limited or archival sources land in the right octave without tuning. - enroll(..., fmax=180) overrides it if a voice still reads the wrong register. Enrollment embeds an input audio sample via…

Open weights apache-2.0 10M parameters
View model

Model · Text to speech

Kokoro-82M

Hexgrad

Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient. With Apache-licensed weights, Kokoro can be deployed anywhere from production environments to personal projects. You can run this basic cell on Google Colab. Listen to samples. For more languages and details, see Advanced Usage. Under the hood, kokoro uses misaki, a G2P library at https://github.com/hexgrad/misaki Model SHA256 Hash: 496dba118d1a58f5f3db2efc88dbdc216e0483fc89fe6e47ee1f2c53f18ad1e4 Data: Kokoro was trained exclusively on permissive/non-copyrighted audio data and IPA…

Open weights apache-2.0
View model