Features

■   Support multiple voice types

      Including clear standard Mandarin, dialects (such as Cantonese and Sichuan dialect), foreign languages (such as English and Japanese), dialects, foreign languages, and real-time transcription need to be customized.

■   Semantic optimization and formatting standardization

      Add commas, periods, question marks, etc. based on tone pauses (such as "um" and "ah") and semantic logic to make the text more fluent.

■   Role differentiation and identification

      In telephone conversation scenarios, automatically distinguish between the two parties and label them (such as "Speaker 1" and "Speaker 2") to clearly present the logic of the conversation.

■   High accuracy of speech recognition

      The accuracy rate of standard Mandarin transcription is over 98%, and the accuracy rate of standard English recognition is over 95%.

■   Supports multiple recording file formats

      Supports audio formats such as pcm/wav/opus/mp3/mp4/m4a/amr/3gp/aac.

■   Adapt to complex environments

      By using noise reduction algorithms, the impact of background noise (such as conference room noise, street noise) and echo on transcription can be reduced, and the recognition accuracy of fuzzy speech can be improved.

■   Support private deployment

      Support private deployment, minimum server configuration is:

      CPU: clock speed above 3.0, 8-core 16 thread, memory: 16GB DDR4, operating system: Linux, domestic operating system.