Review of commercially available reading software 'VOICE PEAK' that AI reads out sentences emotionally with 7 types of voices

AH-Software , which provides input character reading software such as VOICEROID series and Synthesizer V series, has released input character reading software ' VOICEPEAK ' that is equipped with AI speech synthesis technology and can easily read sentences on Windows, macOS, etc. Announced that it will be released for Linux on March 11, 2022. VOICEPEAK contains 6 narrators + 1 girl = 7 types of voices in total, and it is also possible to control the emotions of the voices read aloud by AI. I had the opportunity to touch it a little earlier, so I actually talked to VOICEPEAK in various ways.

VOICEPEAK Commercially available 6 narrator set | Product information | AHS (AH-Software)

There is a trial version of VOICEPEAK, but this time I experienced a beta version that is closer to the product version. The screen of VOICEPEAK looks like this.

When I clicked on the screen, the dialogue input screen was displayed.

As I typed the text, the readings and accents appeared at the bottom of the screen almost at the same time.

Click the beginning of the input text, move the cursor, and click the play icon in the top menu, and VOCIE PEAK will read it aloud. Input character reading software may take some time to generate voice after inputting characters, but VOICEPEAK judges reading and accent on the spot, and the voice is played immediately.

There are a total of 7 types of reading voices: 3 female patterns + 3 male patterns + 1 girl pattern. The following movie is where I actually entered the text and read it out with each voice.

When you talk to VOICEPEAK, which reads out sentences with 7 different voices, it looks like this-YouTube

To the right of the input screen is a slide bar that allows you to adjust the reading speed and pitch. Furthermore, the big point of VOICEPEAK is that you can adjust your emotions when reading aloud. By adjusting 'happiness', 'enjoyment', 'anger', and 'sadness' with the slide bar, the pitch and brightness of the voice are automatically adjusted. With this system that can adjust emotions, it is possible to have the same sentence read aloud in a different atmosphere.

Actually, the following is the place where AI read aloud with 100% each of 'happiness', 'enjoyment', 'anger', and 'sadness'. When 'happiness' and 'enjoyment' are 100%, the tone of the voice becomes brighter and the intonation becomes a little louder, while the intonation is suppressed in anger and sadness, and the pronunciation such as the strength of consonants is slightly. I get the impression that there is a difference.

'VOICE PEAK' that AI reads out the text will change the way of reading out by setting emotions-YouTube

VOICEPEAK also has a built-in dictionary function, which makes it possible to understand and prepare readings even for sentences mixed with kanji. However, sometimes AI makes a mistake in reading. For example, the following is a scene where '10 minutes on foot (Tohojupun)' is read as 'Tohojubun'.

The reading can be corrected by re-entering it directly. As a user dictionary function, VOICEPEAK can also add reading and pronunciation of specific words.

And the intonation can also be adjusted by switching between high and low.

You can also fine-tune the intonation pitch by clicking the red frame in the image below.

So, the place where I actually read '

Space Dangerous' with VOICEPEAK is as follows. I am correcting the reading of some Chinese characters and adjusting the emotional parameters, but other than that, I am asked to read the input as it is. When you paste the text, the reading and intonation are prepared in an instant, and with just a few corrections, it will be read out smoothly.

I asked 'VOICE PEAK', which AI reads out the text naturally, to read the copy and paste of 'Space Danger' --YouTube

To export the generated narration, click the book icon in the top menu and select Output.

Select 'Export as single file' and click 'Export' to export the narration in WAV format.

For the VOICEROID series released by the same AH-Software, it was necessary to purchase a commercial license separately for commercial use. However, VOICEPEAK is commercially available from the beginning. VOICEPEAK can be said to be software that can be expected to be ready for action because AI automatically prepares reading and intonation, narration is generated instantly, and loading does not take time. The narration is also very high quality, and it reads aloud so smoothly that you can't think it's being read aloud by AI. It is also a big feature that it can correspond to dialects if the intonation is set in detail. In addition, there is no need for complicated operations, and the UI is simple, easy to see, and intuitive to use, so it is a nice point that even non-experts can use it easily.

VOICEPEAK is scheduled to be released on March 11, 2022 (Friday), and includes 6 types of narrator + 1 voice, a total of 7 types of voices, and the price is 19,800 yen including tax.

VOICEPEAK Commercially available 6 narrator set | Product information | AHS (AH-Software)

in Review,   Software,   Video, Posted by log1i_yk