Free open-source 2B Chinese speech model reproduces Mangzhuang Ren with high-speed tonguetwisters

国产免费2B开源语音模型征服《莽撞人》！复刻郭德纲最难贯口

ModelBest, OpenBMB, and Tsinghua University released VoxCPM 2, a 2B open speech model that supports 9 Chinese dialects, 30 foreign languages, and 48kHz audio. The post says generation often finishes within 1 second, recommends reference audio of at least 5 seconds, and supports denoising, LoRA, and full fine-tuning; the key detail is its tokenizer-free diffusion autoregressive continuous representation design.

Why it matters: This is a substantive open-source speech release, not a thin demo: the post gives 2B, 48kHz, 9 Chinese dialects, 30 languages, ref audio ≥5s, and a tokenizer-free route. HKR-H/K/R all pass, but the event is not large enough for a must-write P1.