Introduction

AmiVoice API is a speech recognition API that converts speech to text. When you send audio, it returns the spoken content as text. You can create speech-enabled applications such as transcription of meetings or voice dialogue systems.

Documentation Structure

Please see the "Introduction and Operation Guide" for security and operational information before implementation, the "Development Guide" for implementation details, the "Reference" for API specifications, and the "Help" section if you encounter any issues.

📄️ Introduction and Operation Guide

Summarizes information necessary for security, compliance, and operation.

📄️ Development Guide

Explains detailed information necessary for development, such as how to use the API according to your purpose, requests, and responses.

📄️ Reference

API Reference

📄️ Help

Troubleshooting and how to make inquiries

Quick Start

1

Obtain an APPKEY

Register from the User Registration Page and note down the APPKEY displayed in the [Connection Information] on your MyPage. Set it as an environment variable using the following command.

macOS / Linux

export APPKEY=your_appkey_here

Windows (PowerShell)

$env:APPKEY = "your_appkey_here"

Windows (Command Prompt)

set APPKEY=your_appkey_here

tip

The AmiVoice Tech Blog provides step-by-step instructions on how to register as a user and convert an audio file to text using AmiVoice API. Please see the following for reference:

Let's Try Using AmiVoice API

2

Prepare an Audio File

Prepare the audio file you want to transcribe. You can use the following sample audio (test.wav) as is.

For information on supported audio file formats, please see Audio Formats.

3

Execute Speech Recognition

Execute the following. Replace test.wav with the path to your audio file.

curl (macOS / Linux)

curl https://acp-api.amivoice.com/v1/recognize \
     -F d=-a-general \
     -F u=$APPKEY \
     -F a=@test.wav | jq

note

• If the curl command is not installed, download the package for your OS from https://curl.se/ or use a package manager to install curl.
• The result text is Unicode-escaped. In the above command, jq is used to format the response for better readability. If jq is not installed, try executing without the | jq part. The jq command can be downloaded from https://stedolan.github.io/jq/ for your OS or installed using a package manager.

curl (Windows PowerShell)

curl.exe https://acp-api.amivoice.com/v1/recognize `
     -F d=-a-general `
     -F u=$env:APPKEY `
     -F a=@test.wav | jq

note

• In PowerShell, curl is an alias for Invoke-WebRequest, so please specify curl.exe explicitly. Windows 10 version 1803 and later includes curl.exe by default. If it's not included, please install it from https://curl.se/.
• The result text is Unicode-escaped. In the above command, jq is used to format the response for better readability. If jq is not installed, try executing without the | jq part. The jq command can be downloaded from https://stedolan.github.io/jq/ for your OS or installed using a package manager.

curl (Windows Command Prompt)

curl https://acp-api.amivoice.com/v1/recognize ^
     -F d=-a-general ^
     -F u=%APPKEY% ^
     -F a=@test.wav

note

• Windows 10 version 1803 and later includes curl by default. If it's not included, please install it from https://curl.se/.
• The result text is Unicode-escaped. In the above command, jq is used to format the response for better readability. If jq is not installed, try executing without the | jq part. The jq command can be downloaded from https://stedolan.github.io/jq/ for your OS or installed using a package manager.

Python

import os
import requests

with open("test.wav", "rb") as f:
    response = requests.post(
        "https://acp-api.amivoice.com/v1/recognize",
        data={"d": "-a-general", "u": os.environ["APPKEY"]},
        files={"a": f}
    )
    data = response.json()  # The JSON parser automatically converts Unicode escapes to Japanese
    print(data)

4

Check the Results

If successful, a JSON like the following will be returned. The transcription result is included in the text field.

{
  "results": [
    {
      "tokens": [ ... ],
      "confidence": 0.998,
      "starttime": 250,
      "endtime": 8794,
      "text": "アドバンスト・メディアは、人と機械との自然なコミュニケーションを実現し、豊かな未来を創造していくことを目指します。"
    }
  ],
  "utteranceid": "20220602/14/018122d637320a301bc194c9_20220602_141433",
  "text": "アドバンスト・メディアは、人と機械との自然なコミュニケーションを実現し、豊かな未来を創造していくことを目指します。",
  "code": "",
  "message": ""
}

For details on the response content, please see Speech Recognition Result Format.

Next Steps

The Quick Start used the Synchronous HTTP interface. If you want to handle real-time audio sources, you can use the WebSocket interface, and if you want to process large audio files exceeding 15MB, you can use the Asynchronous HTTP interface. For each use case and points to consider when choosing an interface, please see Interface Type and Usage.

📄️ Synchronous HTTP Interface

Simple implementation, optimal for short audio files

📄️ WebSocket Interface

Streaming

📄️ Asynchronous HTTP Interface

Large files and batch processing

We also provide client libraries and sample programs to support development.

📄️ Client Libraries

📄️ Sample Programs

For improving speech recognition accuracy, you can utilize the following features:

📄️ Speech Recognition Engines

You can change the speech recognition engine based on the domain

📄️ User Dictionary

You can pre-register technical terms and proper nouns

📄️ Rule Grammar

Improves accuracy by limiting patterns

We also provide additional features such as speaker diarization and sentiment analysis. Please use them according to your purpose.

📄️ Speaker Diarization

Separates audio containing multiple speakers by speaker and identifies who spoke when.

📄️ Sentiment Analysis

Analyzes emotions from speech and can identify the speaker's emotional state.

We also provide features to support the operation of your built services.

📄️ Usage Aggregation Tags

You can aggregate usage time for each tag set during requests.

Please also see the comprehensive Development Guide.