VO

VoxCPM2

VoxcpmGGUF by DennisHuang648Updated 2026-06-30
Download on Hugging Face
Parameters912M
Context32K
Layers28
Architectureminicpm4
Smallest quant1.7 GB
Quick answer

VoxCPM2 needs about 2.6 GB of memory at F16 (a 1.8 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 1.7 GB (Q8_0) to 3.2 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 0.9 GB.

VoxCPM2 – 0.912B parameter, 48kHz multilingual text‑to‑speech

VoxCPM2 is a tokenizer‑free, diffusion autoregressive TTS model with 0.912B parameters. It supports 30 languages and produces 48kHz mono audio. The model is fully open‑source under Apache‑2.0 and can run on CPU, Metal, CUDA, or Vulkan via llama.cpp‑omni.

Good at

  • Multilingual TTS without language tags
  • Voice design from natural‑language description
  • Real‑time streaming on modern GPUs

Limits

  • No vision support
  • Requires 8k context for memory estimates
  • No explicit language list beyond 30 languages

Best for

  • Creating multilingual synthetic voices
  • Voice cloning from short clips
  • Real‑time TTS applications

Who it’s for. Developers building multilingual TTS services or voice cloning tools. Researchers needing a lightweight, open‑source model for experimentation. Companies seeking a commercial‑ready, Apache‑2.0 licensed solution.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q8_01.7 GB 2.5 GB
Fits Get file →
F161.8 GB 2.6 GB
Fits Get file →
F163.2 GB 4.0 GB
Fits Get file →

For 8 GB VRAM, use Q8_0 (1.73 GB, ~2.5 GB RAM). For 12 GB VRAM, F16 (1.83 GB, ~2.6 GB RAM) is suitable. If more VRAM is available, the larger F16 (3.25 GB, ~4 GB RAM) provides higher quality. Memory estimates are approximate.

What GPU do you need for VoxCPM2?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is F16 (3.2 GB file, about 4.0 GB in use).

Which VoxCPM2 quant should you choose?

F16 (1.8 GB) is the middle option in this repository. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.

If memory is tight, the smallest file is Q8_0 at 1.7 GB, with the quality trade-offs that come with fewer bits. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureminicpm4
Parameters912M
Layers28
Trained context32,768 tokens
KV cache (fp16)28 KB per token

Source

Quantized byDennisHuang648
FamilyVoxcpm
Base modelopenbmb/VoxCPM2
Licenseapache-2.0
Downloads (HF)25,669
Last updated2026-06-30

Frequently asked questions about VoxCPM2

How much VRAM does VoxCPM2 need?

About 2.6 GB at F16 with an 8K context window. That is the 1.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run VoxCPM2 on an 8 GB GPU?

Yes, with F16: the 3.2 GB file needs about 4.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of VoxCPM2 GGUF?

It depends on the quant: Q8_0 is 1.7 GB, F16 is 1.8 GB, F16 is 3.2 GB.

What context length does VoxCPM2 support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.9 GB at 32K.

Where can I download VoxCPM2 GGUF?

The files are on Hugging Face in the DennisHuang648/VoxCPM2-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does VoxCPM2 use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run VoxCPM2 on 8 GB VRAM?

Yes, the Q8_0 quantized model is 1.73 GB and requires about 2.5 GB RAM at 8k context.

Does VoxCPM2 support voice cloning?

Yes, it supports controllable cloning from short clips and ultimate cloning with reference audio and transcript.

Is VoxCPM2 licensed for commercial use?

Yes, it is released under the Apache‑2.0 license, which allows commercial use.

Other GGUFs of VoxCPM2

Similar models

Browse all →

More from DennisHuang648

All by DennisHuang648 →

More Voxcpm models

All Voxcpm →

Other tiny models

All tiny models →