Primary sourceIT之家
View sourceEvent ·Multi-source
Block 推出全双工语音模型 Dessa:可本地运行,原生支持情绪表达
Twitter 联合创始人 Jack Dorsey 旗下的 Block 公司推出语音模型 Dessa,与 OpenAI 的 GPT-Live 一样属于全双工模型,能边说话边收听用户语音。官方称训练仅用了一组英伟达 GPU 机架,且完全基于人类语音数据,情绪表达为模型原生内置,能模拟笑声、停顿、哽咽等状态。该模型可在消费级电脑本地运行,门槛为 M4 Max 及以上配置的 MacBook。目前不对外开放试用,Block 正邀请安全研究者测试语音智能体欺骗人类的风险场景。
Current read
Block 发了个叫 Dessa 的语音模型,能边说边听、自带情绪,哽咽疲惫都能演,M4 Max 以上 MacBook 就能本地跑,目前只给安全研究者测。
11 sources62° AI heatLast updated 2026/10/25 02:29:30
Evidence
11 sourcesRelated source 1AIGCLINK
View sourceRelated source 2TestingCatalog
View sourceRelated source 3Decoder
View sourceRelated source 4Artificial Analysis
View sourceRelated source 5Noam Brown (OpenAI 推理)
View sourceRelated source 6VnExpress Số hóa
View sourceRelated source 7rohanpaul_ai
View sourceRelated source 8Bindu Reddy
View sourceRelated source 9kimmonismus
View sourceRelated source 10karminski-牙医 (AI工具)
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.