就在今天，模思智能及 OpenMOSS 团队再度上新，发布并开源了 MOSS-TTS Family ，一套面向 高保真、高表现力与复杂场景生成 的语音生成模型家族。 你可以用 MOSS-TTS Family 完成这些事情： 从这些真实、明确的实际需求，我们不难看出，模思推出的 TTS 全家桶，并不是单一能力的堆叠，而是一整套 可以直接接入创作流程、产品系统与交互场景的声 音生产工具链 。 语音生成模型家族：全维度能力覆盖 MOSS-TTS Family 并不是对 "一个更大的 TTS 模型" 的追求。 相反，我们选择将声音生产拆解为多个真实存在的创作与应用环节，并为每一个环节提供专门的模型支持，使它们既可以独立使用，也可以组合成完整的工作 流。 整个模型家族包含五个核心成员： 当一段语音不仅需要 "像某个人"、"准确地读出每个字"， 还需要在不同内容中自然切换说话方式， 在几十分钟的叙述中持续稳定， 在对话、角色、实时交互等不 同形态下都能直接使用 —— 单一的 TTS 模型，往往已经不够用了。 它们共同构成了一个 覆盖 "稳定生成、灵活设计、复杂对话、情境补全、实时交互" 的声音创作生态闭环 。 MOSS- ...

Core Viewpoint - Qwen has launched Qwen3-TTS, an open-source voice generation system that supports voice cloning, voice creation, human-like voice generation, and voice control based on natural language descriptions [1] Group 1: Product Features - Qwen3-TTS includes a full series of multi-codebook models that are now open-sourced [1] - The system offers two model sizes: 1.7 billion parameters and 0.6 billion parameters [1] - It supports 10 major languages including Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian, along with various dialects [1]

Core Insights - The company Mianbi Intelligent has released a voice generation base model called VoxCPM with 0.5 billion parameters [1] - VoxCPM was developed in collaboration with THU Shenzhen International Graduate School's Human-Machine Voice Interaction Laboratory [1] - The model achieves state-of-the-art (SOTA) levels in naturalness of synthesized speech, similarity of tone, and prosodic expressiveness [1] - VoxCPM is currently open-sourced on platforms such as GitHub and Hugging Face [1]