Comparison of Chinese Speech-to-Text Accuracy: MeetingMinutes vs. Otter — A Real-World Meeting Test

As a product manager at an internet company, weekly meetings take up nearly half of my working hours. To improve meeting note-taking efficiency, I have tried many recording and transcription tools. Recently, I conducted a comparative test on Chinese transcription accuracy between MeetingMinutes and Otter. Below are my test results.

This test simulated a real office scenario in a standard conference room of approximately 20 square meters. The phone was placed in the center of the conference table, about 3 meters away from the main speaker. The test content was a real product review meeting lasting 32 minutes, with 3 speakers (2 male, 1 female) speaking at a moderate pace in standard Mandarin. The content covered product requirement discussions, KPI reporting, and technical solution reviews.

The test results show that MeetingMinutes achieved a transcription accuracy rate of 98% in standard Mandarin scenarios, with overall recognition performance being highly satisfactory. Its AI filtering function automatically removes common filler words such as "um," "ah," and "like," as well as repetitive expressions and pauses, producing clean and well-structured transcripts that can be copied and used directly without extensive editing.

Otter, as a well-known meeting transcription tool in the international market, achieved a transcription accuracy rate of approximately 92% in this Chinese test. While it performs reasonably well in recognizing everyday vocabulary, there is still room for improvement in recognizing Chinese technical terminology, numerical expressions, and idiomatic expressions. Some sentences also showed deviations in punctuation and semantic understanding.

In terms of user experience, transcripts generated by MeetingMinutes can be directly edited and exported, with an operation flow more aligned with domestic user habits. Otter, on the other hand, is more suitable for English meeting scenarios, with mature English recognition capabilities and cross-device synchronization.