As of today, August 12, 2026, we have made the "Sentiment Analysis Option" available in the Asynchronous HTTP Interface v2. Previously, the Sentiment Analysis Option was only available in the Asynchronous HTTP Interface v1, but it is now also available in v2.

How to Migrate to v2

To migrate to v2, change the destination of your speech recognition requests as follows:

Before change (v1)

https://acp-api-async.amivoice.com/v1/recognitions

After the change (v2)

https://acp-api-async.amivoice.com/v2/recognitions

Specifying the sentiment analysis option is the same as in v1. Specify the following value for the 'd' parameter in the speech recognition request.

sentimentAnalysis=True

Differences Between v1 and v2

In v2, by splitting the audio and processing it in parallel, processing can now be completed in a shorter time than in v1, especially for long audio clips. There is no difference in the accuracy of speech recognition.

While there are no major changes to the basic usage, the response methods and error return methods for some requests will differ. When migrating, please refer to the "About the New Version" section in the manual below.

AmiVoice API Manual - Asynchronous HTTP Interface - About the New Version

About the Sentiment Analysis Option

For details on how to specify parameters for the sentiment analysis option, response specifications, and the content of sentiment parameters, please refer to the following manual.

AmiVoice API Manual - Sentiment Analysis

Important Notice for Customers Using Sentiment Analysis Option on v1

v1 and v2 use different versions of the sentiment analysis engine, Therefore, sentiment parameter values may differ even when processing identical audio content. This is due to the update of the sentiment analysis engine, and there is no change in the definition of the emotions that the parameters represent.

While the parameter specification method and overall response structure are basically the same, if the sentiment parameter is judged by comparing it to a threshold, the judgment results may change when migrating to v2. Therefore, customers using v1 sentiment analysis should process their actual audio in both v1 and v2 before migrating to v2 and check the following.

• Changes in the values ​​of the emotion parameters being used
• The validity of the currently set threshold.

v1 can be used in parallel with v2 until the scheduled end-of-service date. During this period, please conduct comparative testing and adjust thresholds and judgment logic as needed before migrating to v2.

Discontinuation of Asynchronous HTTP Interface v1

The Asynchronous HTTP interface v1 will be discontinued on December 31, 2026.

Customers currently using v1 are requested to migrate to v2 before the service ends.

We appreciate your continued support of the AmiVoice API.