VoxCPM：0.5B参数语音生成模型，实现零样本语音克隆与行业领先自然度
BLUFVoxCPM is a 0.5B parameter speech generation model achieving industry-leading naturalness and zero-shot voice cloning with 0.17 RTF efficiency, supporting Chinese/English synthesis.
AI大模型2026/1/20
阅读全文 →标签：人工智能
查看包含 人工智能 标签的所有文章。
共 1070 篇
LEANN AI框架2024指南：存储效率提升50倍的向量搜索技术
BLUFLEANN AI框架通过动态重算嵌入和压缩邻近图索引，显著降低向量搜索的存储开销，解决传统索引部署难题。
原文翻译：
The LEANN AI framework significantly reduces vector search storage overhead through dynamic embedding recalculation and compressed proximity graph indices, addressing traditional index deployment challenges.
AI大模型2026/1/20
阅读全文 →
VoxCPM-1.5：开源中文TTS模型，3秒克隆人声，消费级GPU高效运行
BLUFVoxCPM-1.5 is an open-source Chinese TTS model with 500M parameters, supporting 44.1kHz audio and 3-second voice cloning. It runs efficiently on consumer GPUs like RTX 3060.
AI大模型2026/1/20
阅读全文 →
LEANN AI框架：向量数据库的SQLite时刻，本地AI搜索新范式
BLUFLEANN is an embedded vector database framework that reduces storage by 97% vs traditional systems while maintaining full accuracy, enabling local AI search without cloud dependency.
AI大模型2026/1/20
阅读全文 →
VoxCPM：0.5B端到端TTS模型，中文语音克隆与本地部署新突破
BLUFVoxCPM is a 0.5B parameter end-to-end TTS model excelling in Chinese voice cloning, symbol pronunciation, and polyphone handling with efficient local deployment capabilities.
AI大模型2026/1/20
阅读全文 →
LEANN AI框架：2024向量搜索存储与计算权衡技术指南
BLUFLEANN AI框架提出一种低存储向量索引，通过创新的存储-计算权衡设计，将索引存储降至原数据5%以下，实现高达50倍的压缩比，为高维向量搜索提供高效解决方案。
原文翻译：
The LEANN AI framework proposes a low-storage vector index that, through an innovative storage-computation trade-off design, reduces index storage to less than 5% of the original data, achieving up to 50x compression ratio, offering an efficient solution for high-dimensional vector search.
AI大模型2026/1/20
阅读全文 →
VoxCPM开源TTS模型架构解析与2024技术指南
BLUFVoxCPM 是一款先进的开源TTS模型，采用无分词器架构与上下文感知技术，支持零样本语音克隆，并在消费级硬件上实现高效实时合成。
原文翻译：
VoxCPM is an advanced open-source TTS model featuring a tokenizer-free architecture and context-aware generation. It supports zero-shot voice cloning and enables efficient real-time synthesis on consumer-grade hardware.
AI大模型2026/1/20
阅读全文 →
LEANN AI框架：向量数据库的SQLite时刻，本地部署RAG存储减少97%
BLUFLEANN is an embedded vector database for RAG that reduces storage by 97% vs traditional solutions while maintaining accuracy, enabling local deployment on laptops with full privacy and no cloud dependencies.
AI大模型2026/1/20
阅读全文 →
VoxCPM：无标记器TTS技术，实现上下文感知语音生成与零样本语音克隆
BLUFVoxCPM is a tokenizer-free TTS system using continuous speech modeling for context-aware generation and zero-shot voice cloning, achieving RTF as low as 0.15 on consumer GPUs.
AI大模型2026/1/20
阅读全文 →