开源语音识别模型:不到 1B 参数可跑在手机上
一个不到 1B 参数、支持 40 多种语言的开源语音识别模型,原生支持 Hugging Face Transformers 和 vLLM,能直接跑在手机上。想做端侧语音转写或离线识别的开发者可以关注,部署门槛很低。
一个全新开源语音识别模型发布,参数量不到 1B(10 亿),支持 40 多种语言,体积和速度足以直接在手机端运行。
模型原生适配 Hugging Face Transformers 和 vLLM 两个主流推理框架,开发者无需额外适配就能部署到本地或服务端。作者展示了它在真实电话录音上的处理效果,说明小模型在嘈杂真实场景下也能保持可用的识别质量。
对需要在端侧做语音转写、通话记录或离线识别的团队来说,<1B 的体积意味着可以在不依赖云端 API 的前提下落地,与 Whisper 等更大的开源方案相比牺牲了一些规模但换来了速度和部署灵活性。
This is an open speech recognition model that's small enough to run on your phone.
It includes native support for Hugging Face Transformers and vLLM.
• Fewer than 1B parameters • 40+ languages • Super fast and small
Even though it is a small model, look at the results processing real phone calls.
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include open-source models, speech recognition, Hugging Face, vLLM.
Source attribution: curated from Ate-a-Pi (@svpino).