Does a voice recorder that uses AI make the minutes to create a minutes?
IFLYTEK's AI voice recorder "VOITER" (Model: SR502J)
Before trial, let's start with the basic specifications.The appearance is a slightly thicker smartphone.In fact, the right side has a SIM slot and also supports Wi-Fi.By using these to connect to the network, you will be able to convert characters in real time.In other words, the recorded sound is immediately textured.After that, by uploading the audio file to a dedicated server, AI will increase the accuracy of the character.However, since audio data is inevitably large, it is better to use Wi-Fi rather than mobile communication in situations where there is a connection environment.
The upper part has a directional microphone and an optimal microphone on the front surface, so that the audio of 360 degrees around a maximum distance of 10 meters can be picked up properly.
The back is equipped with 8 million pixels.Not only photos but also videos are supported, and subtitles are possible.If you are connected to the network, you can even display subtitles in real time in the recorded video.
And one of the problems when creating text from audio automatically is the surrounding noise.Such noise often makes the correct words cannot be determined, so the output characters often go wrong.By the way, VOITER is said to be the "chime", an international contest for voice recognition technology held every other year, with the technology that has been ranked first in three consecutive times from 2016 to 2020.This will make it quite accurate in a noisy place.This can be expected to be finished in a higher accurate sentence by adding the function of the AI correction on the cloud here.
In addition, modes such as "meeting", "lecture", "coverage", "standard", "memo", and "music" can be selected, so that recording and perceptions according to the situation can be performed.The audio file stored in the main unit and the text file that have been written can be transmitted to the e -mail address from the operation screen, exporting the USB cable on a PC.At the same time, files can be managed using the cloud space provided by IFLYTEK.
Let's actually try the automatic voice recognition characterization function.The experiment is to place VOITER and smartphones on the desk, read out the sentence of this press release.Punctions and parentheses are not pronounced, but they look as one sentence.The following part reads out.
Voiter mode is "standard".The app that is used on the smartphone side to compare is the "Voice Character Conversion & Sound Detection Notice" provided by Google.
First, I read it from a distance of about 40 cm from the device in a quiet place.
VOITER result
Results of smartphones
It is inevitable that proper nouns such as company names and model names will become katakana or change.Comparing other parts, it can be seen that Voiter is more accurate.Moreover, it is quite convenient to automatically recognize the phrase from the sentence and put it in the right place.
Next, I read out the music without vocal as a noise from a similar distance while flowing at a considerable volume.The results are as follows.
音楽を流しながらのVOITER result
音楽を流しながらのResults of smartphones
As expected, both aircraft have a misalignment, but VOITER still puts a phrase point and is much closer to the original text.The "personality Momoko" and "Ocean Solo Corner" by smartphones are already in this state.Both must be finished, but smartphones are more difficult.
By the way, the process after uploading will end within 24 hours after uploading.However, "acceleration processing" that is given priority processing can be used, and in this case, it will be completed within 20 minutes.At present, the server is vacant, and the processing process has been completed in a few minutes.
As far as the results of this VOITER are seen, if this character is highly accurate, it will be quite easy to create minutes.Even as our profession, it is an impression that practical accuracy is given during interviews such as interviews.
Currently, VOITER is sold on this project page of the crowdfunding site "Makuake".The price -included price is 43,800 yen for pre -sale.After all, it is a bit high for high functionality, but I would like to evaluate that I will be able to play an active role just to match it.
Finally, if you summarize the detailed specifications of Voiter, the storage is 16GB, the record format is WAV and AAC, the video is MP4, the text is TXT, and the image is JPG.The camera is 8 million pixels, and the compatible language is Japanese and English.Data communication supports 4G, Wi-Fi and Bluetooth.The body size is W62.26 x D125 x H14.8mm and weighs 166g.
The more compact "VOITER MINI" is also sold, and this is the lowest price of Makuake and the tax -included price is 16,280 yen.The body size is W32 x D55 x 15mm and weighs 30g.