# VoxCPM2: Open-source 2B-parameter multilingual TTS with voice cloning hits GitHub Trending

> **Key takeaway:** OpenBMB released VoxCPM2, a 2-billion-parameter tokenizer-free text-to-speech model supporting 30 languages with voice design, controllable cloning, and 48kHz output under Apache-2.0 license.

**Published:** 2026-10-24T06:02:33.510809+00:00 
**Source:** GitHub Trending (daily) 
**Category:** foss-apps 
**Canonical URL:** https://fosswire.org/news/voxcpm2-open-source-2b-parameter-multilingual-tts-with-voice.html 

## Summary
VoxCPM2 is a new open-source text-to-speech system from OpenBMB built on the MiniCPM-4 backbone. The 2B-parameter model was trained on over 2 million hours of multilingual speech data and supports 30 languages including Turkish. It offers three voice generation modes: Voice Design from text descriptions, Controllable Voice Cloning from reference audio with style adjustments, and Ultimate Cloning that continues from reference audio with transcript matching. The model outputs 48kHz audio directly without external upsamplers and can achieve real-time factors as low as 0.13 with vLLM-Omni acceleration. Weights and code are released under Apache-2.0 for commercial use.

## What VoxCPM2 offers technically
- **Tokenizer-free architecture**: Uses end-to-end diffusion autoregressive design that generates continuous speech representations directly, avoiding discrete tokenization steps
- **Three generation modes**: Voice Design (text-to-voice), Controllable Cloning (reference audio + style guidance), Ultimate Cloning (reference audio + transcript continuation)
- **Audio quality**: 48kHz output via AudioVAE V2 asymmetric encode/decode with built-in super-resolution; accepts 16kHz reference input
- **Inference performance**: RTF ~0.3 on RTX 4090, ~0.13 with Nano-vLLM or vLLM-Omni serving with PagedAttention and OpenAI-compatible API
- **Deployment options**: Python API, CLI, web demo, production vLLM serving, and on-device inference via llama.cpp-omni
- **Requirements**: Python 3.10-3.12, PyTorch ≥2.5.0, CUDA ≥12.0

## Why it matters
VoxCPM2 enters a competitive open-source TTS landscape as one of the few fully open-weight, commercially-licensed models with voice cloning capabilities. Its vLLM-Omni integration and OpenAI-compatible API lower deployment barriers for developers building voice applications, while the llama.cpp-omni path suggests potential for local-first or edge deployment.