Starting March 26, 2025, a new end-to-end speech recognition engine will be available with the AmiVoice API.

The newly added engines are as follows:

言语	engine name	Supported sampling rates	Connection Engine Name
Japanese	Japanese E2E_general purpose	8k / 16k	-a2-ja-general
Chinese	Chinese E2E_general	8k / 16k	-a2-zh-general
Multilingual	Multilingual E2E_general purpose	8k / 16k	-a2-multi-general
Japanese	Japanese E2E_general batch	8k / 16k	-a2b-ja-general
Chinese	Chinese E2E_general batch	8k / 16k	-a2b-zh-general
Multilingual	Multilingual E2E_general batch	8k / 16k	-a2b-multi-general

The "Japanese" and "Chinese" speech recognition engines can be used for general transcription purposes."Multilingual"The voice recognition engine is capable of transcribing audio containing a mixture of languages ​​into each language. Supported languages ​​are Japanese, English, and Chinese.

In addition, among the end-to-end speech recognition engines,Batch processing engine(Connection engine name ends with b The batch processing engine is suitable for cases where recognition accuracy is more important than real-time performance. For general-purpose applications, you can expect to improve recognition accuracy in many cases by using a batch processing engine.

On the other hand, if it is important to support vocabulary specific to a specific industry, such as the medical field, we recommend that you continue to use the previous engine. Please rest assured that we will continue to work on improving the accuracy of the previous engine as an important mission.

Also, the newly released next-generation engine has some limitations compared to the previous engine. For details, see the documentation.Speech Recognition EngineSee "Difference between End to End and Hybrid" in the

【重要】
To these next-generation enginesIt won't switch automaticallyTo use this feature, you must explicitly specify the "Connection Engine Name" listed in the table above when making an API request.

At AmiVoice Cloud Platform, we are concentrating our resources and preparing to deliver even better services and new value. Please stay tuned.