Shangtang AI Speech Input and Intelligent Agents
SoWork is a voice input tool launched by SenseAudio, a subsidiary of SenseTime. Its focus isn’t on the accuracy of recognition itself, but rather on the subsequent steps that follow the recognition process. Traditional voice input methods involve dictation: whatever you say is directly typed out, and any mistakes have to be corrected later. Phrases repeated throughout the speech, as well as any jumbled or disordered words, are all displayed on the screen. SoWork, however, utilizes large language models for post-processing tasks. It can smooth out spoken language, add punctuation marks, and transform a messy sequence of oral statements into coherent written text.
The product offers both web-based and Windows client access points. The desktop version is the focus here – when using voice commands on mobile devices, the messages are usually short, consisting of just a few words. However, when speaking in front of a computer, the content often includes emails, weekly reports, or meeting minutes, which can contain hundreds or even thousands of words. Whether such lengthy messages can be effectively organized and processed via voice commands is crucial for the feasibility of using this feature.
The target audience is quite clear: individuals who type slowly but speak smoothly and need to produce large amounts of text daily. This tool does not address creative issues; rather, it helps in quickly translating thoughts into written documents. It is currently available for free trial use.