88Z AI, 7 April — OpenBMB has released VoxCPM2, an open-source speech synthesis model with 2 billion parameters. It supports 30 languages and Chinese dialects, delivering 48kHz studio-quality audio. The model introduces Voice Design for creating new voices via text descriptions, alongside controllable and ultimate cloning modes. Trained on 2.36 million hours of data with a real-time factor of 0.13, it is licensed under Apache-2.0 for commercial use.
At a glance
- Supports 30 languages and Voice Design for text-based voice creation
- Trained on 2.36 million hours of data, Apache-2.0 licensed for commercial use
The update
OpenBMB has officially open-sourced VoxCPM2, a speech synthesis model with 2 billion parameters. Built on a tokenizer-free diffusion autoregressive architecture, it supports 30 languages, including Chinese dialects, and outputs studio-quality audio at 48kHz. VoxCPM2 pioneers the Voice Design feature, allowing users to create unique voice characters solely through text descriptions. It also supports controllable voice cloning and an ultimate cloning mode to meet diverse needs. Trained on 2.36 million hours of data, the model achieves a low real-time factor of 0.13 for efficient performance. Released under the Apache-2.0 license, it is available for commercial use.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- VoxCPM2Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.