Tech blog
-
2026.04.24"Speech segment ratio" as seen in operational data
"How much does speech recognition actually cost?" -- To accurately understand the cost of the AmiVoice API, which charges only for the spoken portion, it's crucial to know the actual "speech segment ratio." This article analyzes the distribution of "speech segment ratios" based on real-world operational data. We'll provide concrete figures to give you an idea of the costs for each usage scenario, such as call centers, smartphone apps, and conference recordings. Please use this information to help you estimate costs before implementation.
-
2026.03.27AmiVoice API Update Explanation: New Parameters for Voicebots Reduce Response Wait Times
Are you familiar with the new AmiVoice API parameters, "recognitionTimeout" and "noInputTimeout," developed based on customer feedback? By utilizing these parameters, you can terminate the processing of unnecessary voice data and control the waiting time during periods of silence, significantly improving the response quality of your voicebot. This article explains how each parameter works and provides specific usage examples.
-
2026.02.19AmiVoice API Update: End-to-End ”Keyword Biasing” Feature
This article explains the mechanism and practical use of "word emphasis," which is now available in end-to-end speech recognition. It covers the differences in behavior compared to hybrid speech recognition, the key points of setting parameters, and tips for improving accuracy. It's a must-read for anyone who wants to get the most out of E2E.
-
2026.01.16AmiVoice's word registration API gives you more freedom in speech recognition!
One of the features of AmiVoiceAPI is the "word registration function." Did you know there is an API for this function? The word registration API allows for flexible word registration, including integration into apps and management of multiple profiles (dictionaries). This article provides an easy-to-understand explanation of how to use the word registration API and some useful scenarios.
-
2024.04.03Why choose AmiVoice API?
We have explained why AmiVoice API is chosen. There are many advantages over other companies' APIs, and we will explain each one one by one. You can also download a comparison table of APIs with competitors that we created, so please check it out.
-
2023.09.29[Comparative verification using the same utterance] Differences in recognition results between Voice Input engines and Conversation engines
We will explain the features of engines that use the AmiVoice API's acoustic model for voice input and engines that use the acoustic model for conversation input, as well as the usage scenarios that each is suited to.
-
2023.07.31Accuracy verification included! Introducing a speech recognition engine specialized for specific applications
We will introduce the voice recognition engine provided by AmiVoice API Private, which is specialized for recognizing names, addresses, etc. We also conducted a verification test comparing its accuracy with a general-purpose engine.
-
2023.07.10What is AmiVoice API Private SDK's "Rule Grammar" Recognition?
We will explain the differences between the "Rule Grammar" speech recognition engine provided by AmiVoice API Private and regular dictation recognition, the appropriate usage scenarios, and the advantages and disadvantages of using it.
-
2023.05.15How to choose whether to display or remove unnecessary words (fillers) with AmiVoice API
AmiVoiceAPI has a function to automatically remove unnecessary words (fillers). However, depending on the situation when using speech recognition, it may be better to display the fillers without removing them. This time, we will explain how to control filler removal.
-
2023.04.17How to convert stereo audio files into two mono audio files
When using AmiVoice API to recognize a stereo audio file, only one channel is recognized. Assuming that the right and left channels of a stereo audio file contain different sounds, this article explains how to convert a stereo audio file into two mono audio files using a tool called SoX.
-
2023.04.03Where does the voice data go after it is processed?
We will answer security-related questions frequently asked by customers considering using the AmiVoice API, such as how voice recognition processed audio is managed, and which plan to choose if the audio data contains personal information.
-
2022.12.20[For intermediate users] About automatic conversion of word spoken in AmiVoice
In Japanese, the reading and actual pronunciation can sometimes vary slightly, such as when "teacher (sensei)" is sometimes pronounced "sensei." AmiVoice automatically changes the reading you specify to accommodate these variations, but if you want to specify the reading precisely, it can take a little skill. This time, we will explain in detail how to specify the reading.
Most viewed articles
-
A quick explanation of how speech recognition works!
-
Comparing the speech recognition rates of OpenAI's Whisper and AmiVoice for "conference" audio
-
How to use the AmiVoice API free coupon
New articles
-
How to use coupons for Zenn Spring 2026
-
"Speech segment ratio" as seen in operational data
-
AmiVoice API Update Explanation: New Parameters for Voicebots Reduce Response Wait Times
Category list
- Introduction to Speech Recognition (15)
- How to improve voice recognition accuracy (12)
- I tried developing it (27)
- How to use AmiVoiceAPI(27)
- Comparison and Verification (6)
- Others(10)
Archive
