An INT8-quantized repack of openbmb/VoxCPM2, prepared so the model can be loaded and run entirely inside a web browser by shi3z/voxcpm2web (a WASM + WebGPU port of Nothing about the model architecture was changed. This repository exists to solve 1. wasm32 has a 4 GB address space. The upstream checkpoint is 4.27 GiB of weights plus a 359 MB audiovae.pth. 2. Quantizing during load costs VRAM you may not have. Doing it ahead of time means the browser never materializes the full-precision tensors at all. There is no separate audiovae file: the AudioVAE weights are already merged in, so a browser needs exactly one large download instead of two. Both rows produce applied=632 missing=0 unused=0…
VoxCPM2 is an open-weight model for text to speech from OpenBMB, released under Apache License 2.0. It has 2.3B parameters. At 16-bit it needs about 5.5 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 385.5k downloads a month.
VoxCPM2 is a tokenizer-free, diffusion autoregressive Text-to-Speech model — 2B parameters, 30 languages, 48kHz audio output, trained on over 2 million hours of multilingual speech data.
Parameters2.3B
Context—
Weights5.0 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads385.5k
Runs On
What it takes to serve VoxCPM2 (2.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 4.6 GB | 5.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 2.3 GB | 2.7 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 1.1 GB | 1.4 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 Oct 14, 2026.
VoxCPM2 on every accelerator the SAVRN Index prices, at every precision
SAVRN's Notes on VoxCPM2
Where does a 2.3 billion parameter text-to-speech model belong in a facility built around 192 GB accelerators? Beside something bigger. VoxCPM2 needs 5.5 GB at 16-bit, 2.7 GB at 8-bit and 1.4 GB at 4-bit, and the cheapest priced slot, one MI300X at $1.85 an hour, has room for many copies. It outputs 48 kHz audio in 30 languages with no language tag, designs a voice from a written description, and clones one from a short clip.
Apache 2.0 keeps the deployment conversation short: commercial use, modification and redistribution are permitted, provided the license, copyright notices and any NOTICE file stay with the files and significant changes are stated. Two checks before committing: the runtime is its own voxcpm library rather than transformers, so plan the serving path around it, and no context length is recorded. The paper is arXiv:2509.24650; the files were last updated 2026-08-18.
Model Card
By OpenBMB, published under apache-2.0, revision 32279effe8c1.
VoxCPM2 is a tokenizer-free, diffusion autoregressive Text-to-Speech model — 2B parameters, 30 languages, 48kHz audio output, trained on over 2 million hours of multilingual speech data.
Highlights
Identity and Version
- Repository
- openbmb/VoxCPM2
- Publisher
- OpenBMB
- Task
- Text to speech
- Modality
- Audio
- Library
- voxcpm
- Parameters
- 2.3B parameters
- Languages
- zh, en, ar, my, da, nl, fi, fr
- Revision
- 32279effe8c19989596f05d353d1447f51d9e915
- First published
- 2026-04-03
- Last updated
- 2026-08-18
Files and Weights
9 files, 5.0 GB in total. The weights are 2 files totalling 5.0 GB in pth, safetensors.
Weights2 files · 5.0 GB
Configuration3 files · 8.9 KB
Tokenizer2 files · 3.7 MB
Documentation1 file · 7.9 KB
Repository1 file · 1.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| audiovae.pth | Weights | 377.0 MB | 94b5d51e107e |
| model.safetensors | Weights | 4.6 GB | f7f964cfa9da |
| config.json | Configuration | 4.3 KB | — |
| special_tokens_map.json | Configuration | 1.6 KB | — |
| tokenization_voxcpm2.py | Configuration | 2.9 KB | — |
| README.md | Documentation | 7.9 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
| tokenizer.json | Tokenizer | 3.7 MB | — |
| tokenizer_config.json | Tokenizer | 5.1 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 5.0 GB
Download from OpenBMB
Released by OpenBMB through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:2509.24650
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 5.0 GB |
| 16-bit | 4.6 GB |
| 8-bit | 2.3 GB |
| 4-bit | 1.1 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Built on This Model
- Derived fromVoxCPM2-q8
Questions About VoxCPM2
How much GPU memory does VoxCPM2 need?
About 5.5 GB at 16-bit and 1.4 GB at 4-bit: the weights (2.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run VoxCPM2 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 VoxCPM2 commercially?
Yes. VoxCPM2 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.
Similar Models
MOSS‑TTS Family is an open‑source speech and sound generation model family from MOSI.AI and the OpenMOSS team. It is designed for high‑fidelity, high‑expressiveness, and complex real‑world scenarios, covering stable long‑form speech, multi‑speaker dialogue, voice/character design, environmental sound effects, and real‑time streaming TTS. When a single piece of audio needs to sound like a real person, pronounce every word accurately, switch speaking styles across content, remain stable over tens of minutes, and support dialogue, role‑play, and real‑time interaction, a single TTS model is often not enough. The MOSS‑TTS Family breaks the workflow into five production‑ready models that can be…
DuDE is a full-duplex conversation TTS (text-to-speech) model that generates two speakers together from an XML dialogue. It supports turn-taking, backchannels, and overlapping speech, with a separate voice for each speaker. The output is a 24 kHz stereo WAV: speaker A on the left, speaker B on the right. DuDE first self-distills Qwen3-TTS on independent utterances to adapt a shared backbone to two interleaved channels. It is then fine-tuned on approximately 1,922 hours of synchronized Seamless Interaction conversations to learn turn-taking, backchannels, and overlapping speech. The XML frontend resolves turn dependencies into interleaved text inputs. Reference-speaker embeddings condition…
Qwen3-TTS covers 10 major languages (Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian) as well as multiple dialectal voice profiles to meet global application needs. In addition, the models feature strong contextual understanding, enabling adaptive control of tone, speaking rate, and emotional expression based on instructions and text semantics, and they show markedly improved robustness to noisy input text. Key features: Intelligent Text Understanding and Voice Control: Supports speech generation driven by natural language instructions, allowing for flexible control over multi-dimensional acoustic attributes such as timbre, emotion, and prosody.…
We release Qwen3-TTS, a series of powerful speech generation models developed by Qwen, offering comprehensive support for voice cloning, voice design, ultra-high-quality human-like speech generation, and natural language-based voice control. Qwen3-TTS covers 10 major languages (Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian) as well as multiple dialectal voice profiles. Key features: Install the qwen-tts Python package from PyPI: Zero-shot speech generation on the Seed-TTS test set (Word Error Rate (WER, ↓)): If you find our paper and code useful in your research, please consider giving a star and citation
VibeVoice is a novel framework designed for generating expressive, long-form, multi-speaker conversational audio, such as podcasts, from text. It addresses significant challenges in traditional Text-to-Speech (TTS) systems, particularly in scalability, speaker consistency, and natural turn-taking. A core innovation of VibeVoice is its use of continuous speech tokenizers (Acoustic and Semantic) operating at an ultra-low frame rate of 7.5 Hz. These tokenizers efficiently preserve audio fidelity while significantly boosting computational efficiency for processing long sequences. VibeVoice employs a next-token diffusion framework, leveraging a Large Language Model (LLM) to understand textual…