Skip to main content

Update History

This page lists the major update history for AmiVoice API and this manual. Minor updates and fixes are made regularly in addition to those listed here. You can also receive updates and known issues through the RSS feed (in Japanese only).

October 2026​

2026-10-14: Whether an API key has the "API Key Issuance Feature" can now be checked on MyPage API​

API keys issued on MyPage can be checked in a list on MyPage (Checking Issued API Keys), and you can now also check whether each API key has been granted the "API Key Issuance Feature".

September 2026​

2026-09-30: Released a feature that allows multiple user dictionary profiles to be used simultaneously, and other changes API​

For a single speech recognition request, you can now use up to five user dictionary profiles simultaneously.

For details, please see Using Multiple User Dictionaries Simultaneously in a Single Request.

In addition, the End to End engine has become a new generation, improving accuracy.

  • Compared with the previous generation, the average error improvement rate in internal testing is approximately 9.7% for 16kHz audio and approximately 2.4% for 8kHz audio in the real-time version (-a2-ja-general). In the batch version (-a2b-ja-general), it is approximately 8.6% for 16kHz audio and approximately 2.6% for 8kHz audio.
  • Compared with the hybrid 会話_汎用 engine, the average error improvement rate in internal testing is approximately 4.7% for 16kHz audio and approximately 1.3% for 8kHz audio in the real-time version. In the batch version, it is approximately 20.0% for 16kHz audio and approximately 13.3% for 8kHz audio.
  • The performance of keyword biasing has also been enhanced, making the effect of keyword biasing more apparent.

2026-09-24: Published How to Send Authentication Information in HTTP Request Headers for the Synchronous and Asynchronous HTTP Interfaces Manual​

For the Synchronous HTTP interface and creating jobs with the Asynchronous HTTP interface, we have published how to send authentication information (API key) in the HTTP request header. Instead of the u parameter, specify it in the Authorization header in the format Bearer {API_KEY}.

Location: Synchronous HTTP Interface, Asynchronous HTTP Interface, Request Parameters

2026-09-16: Released a feature to specify the channel to recognize for multi-channel audio data API​

For audio data whose header contains audio format information and that has multiple channels, speech recognition previously targeted only the first channel, but by using the request parameter audioFilterProperties, you can now specify the target of speech recognition. In addition, audio data with three or more channels is now supported. Note that this feature is available only with the Synchronous HTTP interface and v2 of the Asynchronous HTTP interface.

For details, please see audioFilterProperties and Audio Format.

2026-09-16: Published How to Retrieve Statuspage Information via API Manual​

We have published how to retrieve, via an API, the information on the Statuspage where you can check information about incidents and performance degradation of AmiVoice API.

Location: Retrieving Information via API

2026-09-14: Added a feature to check disabled API keys on MyPage, and other changes API​

API keys that were manually disabled on MyPage can now be checked in [Connection Information] on MyPage. For details, please see Checking Issued API Keys.

In addition, the service ID that was displayed in "Connection Information" on MyPage has been changed to be displayed in "Account Information". Note that for accounts created before the MyPage renewal on 2025-06-30, it continues to be displayed in "Connection Information" as well.

August 2026​

AmiVoice API now supports AWS PrivateLink, which allows you to connect within the AWS network without going through the internet. Via an interface endpoint created within your VPC, you can use the speech recognition, user dictionary operation, and API key issuance APIs with the same API key as before. Prior application is required to use this feature.

For details, please see Private Connection (AWS PrivateLink).

2026-08-12: Released the sentiment analysis feature on the Asynchronous HTTP interface v2 API​

The sentiment analysis feature was previously provided only on v1 of the Asynchronous HTTP interface, but it is now available on v2 of the Asynchronous HTTP interface as well.

The usage, the way parameters are specified, and the overall structure of the response are basically the same between v1 and v2, but the version of the sentiment analysis engine differs, so even when processing the same audio, the values of the sentiment parameters will not be identical. This is due to an update of the sentiment analysis engine, and the definitions of the emotions indicated by the parameters have not changed. Please keep this in mind when migrating from v1 to v2.

Note that on v1 of the Asynchronous HTTP interface, you can continue to use the sentiment analysis feature in parallel with v2 until the end of service.

2026-08-04: Published How to Specify Rule Grammar Manual​

We have published how to specify the rule grammar used in speech recognition requests when using the "音声入力_ルール" engine (-a-rule-input-private) available in AmiVoice API Private. Along with this, we have reorganized the structure of the sections that contain information about rule grammar.

Location: How to Specify Grammar Files under Rule Grammar

July 2026​

2026-07-28: Change to MyPage Display (Some Users) API​

For accounts created after the MyPage renewal on 2025-06-30, we have switched the Service ID and Service Password that were displayed in the "Connection Information" on MyPage to be hidden. These were provided for use in authentication in the API Key Issuance API, but from now on, please use an API key issued with the "API Key Issuance Feature" enabled on MyPage.

The Service ID and Service Password cannot be changed unless the account is deleted. Therefore, in order to reduce the risk of leaking authentication information, we have stopped displaying them on MyPage.

Note that for accounts created before the MyPage renewal, they continue to be displayed in the "Connection Information". Since they cannot be changed, please handle them with sufficient care.

2026-07-23: Started detecting partial outages of speech recognition engines on the Statuspage through periodically executed automated tests API​

Until now, we monitored the speech recognition server logs, automatically detected communication errors between internal servers, and displayed them on the Statuspage as "接続障害" (Connection Failure). However, with this method alone, we could not sufficiently detect cases where clients were unable to obtain speech recognition results due to factors other than internal communication errors.

We have now added an E2E test that periodically sends test audio from outside to each speech recognition engine, in the same way as actual use, and confirms that recognition results can be obtained normally. This has strengthened the automatic detection of partial outages.

If a state in which a speech recognition engine does not return recognition results continues for some reason, that situation is automatically detected and displayed on the Statuspage.

We have added security-related information for AmiVoice API, including credit card payment information and the division of responsibility.

Location: Service Specifications

2026-07-10: Guidance for Generative AI and Coding Agents Published Manual​

1. Publication of the Index File llms.txt​

We have published llms.txt, an index file for AI that consolidates links and overviews of the main API specifications and various documents. By having an AI agent read this file first, it can access the information it needs by the shortest path.

2. Provision of Markdown Versions of the Documentation​

For all documentation pages, we now support output in a plain Markdown format that is easy for AI to interpret and considerate of token reduction. It can be obtained by either of the following methods.

  • Request by appending .md to the end of each documentation page URL.
  • Request by adding the Accept: text/markdown header to the HTTP request.

3. Addition of Link Headers to Make the Above Content Easy for AI to Discover​

So that AI crawlers and agents can automatically discover these resources, we add a Link header indicating the URL of the corresponding Markdown version and the location of llms.txt to the HTTP headers of ordinary HTML page responses.

4. Guidance on How Developers Can Make Use of These Features​

We have published guidance on how to make use of the AmiVoice API when carrying out development with generative AI and coding agents by utilizing the new features described above.

Location: Development with Generative AI

2026-07-09: Separated the API Reference from the Manual Manual​

The AmiVoice API reference has been separated from the manual so that it can be viewed from a dedicated tab. For the HTTP interface API, we have migrated to a reference automatically generated from an OpenAPI-format specification. This allows you to always check the latest specification, as well as import the specification directly into various development tools.

Location: Reference

June 2026​

2026-06-26: How to Send Authentication Information in HTTP Request Headers with WebSocket Published Manual​

For the WebSocket interface, we have published a method to send the authentication information (API key) by writing it in the HTTP request header when establishing the WebSocket connection, instead of sending it within the s command after the WebSocket connection.

Location: How to Send Authentication Information in HTTP Request Headers

2026-06-12: "Quick Start" for User Dictionary Published Manual​

A quick start guide has been published to allow easy testing of the user dictionary feature.

Location: User Dictionary

2026-06-08: Terms of Service and SLA Pages Opened Manual​

The terms of service and SLA have been newly published in a separate tab of the manual, allowing access to update history and some past versions.

Location: AmiVoice API Terms of Service / SLA

May 2026​

2026-05-28: Release of New Features and Request Parameters API​

A feature to obtain language labels for word-level recognition results has been released. This can be enabled with useLid in the request parameter "segmenterProperties".

Additionally, the parameter "confidenceLevel" for specifying the confidence threshold for recognition results, and the parameters "completeTimeout" and "wildcardModelPenalty" specific to the "音声入力_ルール" engine have been added.

Furthermore, existing parameters maxResponseTime and targetResponseTime have been modified to be effective only for real-time speech recognition, while maxDecodingRate and targetDecodingRate are now effective only for batch processing of pre-recorded audio data.

2026-05-08: Information About MyPage Published Manual​

Information about the MyPage of AmiVoice API has been published.

Location: MyPage (Management Page) Operations

April 2026​

2026-04-14: Renamed "APPKEY" to "API Key" Manual​

We have changed the name of the "service authentication key string" required for using AmiVoice API from "APPKEY" to "API Key". We have also unified the term "One-time APPKEY" to "API Key". Additionally, we have added an explanation page about API keys and partially changed the structure of the manual.

Location: API Key and its sub-pages

We have also corrected an error in the manual regarding the error message displayed when using an API key issued with incorrect service ID or service password in the API Key Issuance API reference.

Location: The "Note" section in API Key Issuance

2026-04-03: Published Reference Information for Adjusting Speech Detection Request Parameters Manual​

We have published recommended values for different usage scenarios and examples of adjustments for speech detection-related parameters within the segmenterProperties request parameter.

Location: Recommended values in segmenterProperties, examples of parameter adjustments in Adjusting Speech Detection Parameters

March 2026​

2026-03-24: Published Information on Handling User Dictionary Data Manual​

As an element of AmiVoice API's service specifications, we have published information on handling user dictionary data.

Location: Handling of User Dictionary Data and Logging Options and Data Provision in Service Specifications

2026-03-12: Renewed "Getting Started" Page Manual​

We have renewed the "Getting Started" page and included a quick start guide to make it easier to begin using AmiVoice API. The previous tutorials have been integrated into the usage pages for each interface.

Location: Getting Started (Quick Start)

2026-03-11: Published Service Specification Information Manual​

We have published information about AmiVoice API's service specifications, including security aspects.

Location: Service Specifications

2026-03-02: Updated Client Libraries API​

We have released version 1.1.12 of the client libraries, which includes updates to support newly released features of AmiVoice API and newer versions of various programming languages.

February 2026​

2026-02-16: Release of keyword biasing feature for End to End engine API​

The Keyword Biasing for End to End Engine has been released. This feature is similar to the "word registration" in the Hybrid engine and makes it easier to recognize words that are difficult to recognize.

Also, with this release, we have decided to refer to both the "word registration" and "keyword biasing" features collectively as "User Dictionary".

January 2026​

2026-01-08: Added instructions for setting email addresses for failure notifications Manual​

We have added documentation explaining how to set up email addresses for receiving notifications about AmiVoice API failures. Along with this, we have partially changed the structure of the manual.

Location: Setting Up Failure Notification Recipients

December 2025​

Updated sections: "Speaker diarization parameters" in segmenterProperties, Adjusting speaker detection sensitivity in speaker diarization

2025-12-02: Addition of Statuspage information Manual​

Information about Statuspage, which provides information on AmiVoice API outages and delays, has been added.

Added section: Outage and delay information

November 2025​

Multiple new features related to speech recognition processing have been released, and maxDecodingTime, maxResponseTime, maxDecodingRate, targetResponseTime, targetDecodingRate, recognitionTimeout, and noInputTimeout have been added as request parameters to enable these features.

Additionally, the Retrieving Profile List has been released for the user word registration API.

October 2025​

2025-10-31: "Usage Aggregation Tag" now supports Asynchronous HTTP interface v2 API​

Usage Aggregation Tags now supports Asynchronous HTTP interface v2, making it available for all interfaces.

2025-10-02: Language labels added to recognition results of multilingual End to End engine API​

Language labels indicating the language of words recognized by the engine have been added to the word-level recognition results of the multilingual End to End engine. For details, please see results[0].tokens[].language in Word-level Results.

2025-10-01: Information on Rule Grammar published Manual​

Information about grammar files used with the "音声入力_ルール" engine (-a-rule-input-private) available in AmiVoice API Private has been published.

Location: Rule Grammar

September 2025​

2025-09-30: Release of new feature "Usage Aggregation Tag" API​

Usage Aggregation Tags, which enables obtaining usage by attribute when sharing the same account, has been released. However, this feature is not yet supported for Asynchronous HTTP interface v2.

August 2025​

2025-08-13: Support for Opus audio data stored in WebM container API​

AmiVoice API now supports Opus audio data stored in WebM container. For details, please see Audio formats with headers.

2025-08-13: Information on segmenterProperties published Manual​

Information about parameters related to adjusting speech segment detection in speech recognition within the request parameter segmenterProperties has been published.

Location: "Speech segment detection parameters" in segmenterProperties

June 2025​

2025-06-23: Word Registration API now supports 8k audio API​

Word Registration API can now handle operations for 8k audio profiles. For details, please see adf in the "Request Parameters" of each operation description in User Word Registration.