Tokyo Archive Center Co., Ltd. a company that provides “AI Transcription,” a service that converts audio and video data into text using speech recognition technology, has adopted the AmiVoice API. We spoke with Mr. Tanabe, a director of the company, about the background behind this decision and the results achieved.

Product/Service Overview

A Simple Audio and Video Transcription Service

AI transcription can be ordered for each audio or video file. When transcribing, you can try out the first minute for free using three different speech recognition engines: AmiVoice API, Google, and Microsoft Azure, and then purchase after confirming the accuracy.
AmiVoice is overwhelmingly the most popular choice for Japanese*, and we can see the high accuracy of AmiVoice's engine.

※reference:AI Speech Recognition Engine Usage Trend Report - Usage of 10,000 AI Transcriptions

The "AI Transcription" screen. Easy to use, just drag and drop audio or video files.

Key Drivers of the Adoption

High Marks for Accuracy, API Integration Ease, and Responsive Support

Before launching the service, we conducted a comparative evaluation of several speech recognition engines, but decided to adopt the AmiVoice API due to its accuracy in Japanese and ease of API implementation. We were able to implement it with confidence due to its high convenience, including support in Japanese, support for questions, and response times.
We are grateful for the generous support system that has allowed us to continue operating the service with peace of mind, from consultations during development to feature improvements and information provision even after the service launch.

The difficulty of implementation

Client libraries and sample programs in a variety of languages ​​were available, allowing for smooth implementation.
Additionally, the structure of the data returned from the API is excellent, and I feel that it is extremely useful in that it can be utilized in the development of a variety of applications.

Impact of the Implementation and Future Outlook

High Repeat Rate for "Japanese Transcription" Orders

We have noticed a high rate of repeat orders among customers who have used our "Japanese Text Conversion" service.
It appears that our service is primarily used by writers and other people who create their own final manuscripts to reduce their workload. Our service has been well received, as it allows users to check the results immediately, is highly accurate, and gives a rough idea of ​​the audio content.

Looking ahead, if we can achieve highly accurate text conversion using soeech recognition technology, we believe that this will free up human work time to finish the final product, making it possible to create high-quality manuscripts.
We look forward to the future evolution of the AmiVoice API, which achieves high accuracy.

Recognition result screen