Utilizing AI Automatic Voice Tools in Multilingual Localization

2. AI Automatic Voice Utilization in Multilingual Localization

Challenges of human narration and their resolution through AI automatic voice

Narration by humans is essential for commercials and dramatic narrations because it has a delicate and strong persuasive power through voice tone and intonation. However, there are challenges in production. Those who have been involved in production may have experienced this. For example, scheduling adjustments with narrators and recording studios may be necessary, re-recording may not go smoothly when changes are made to the script, or it may be difficult to find narrators in that language in Japan. Of course, if delicate persuasive power is required for narration, it is necessary to create it with human narration. However, in cases where it is not necessary, using AI automatic voice can solve these challenges.

Our company utilizes AI automatic voice tools for the following:

・Operation Instructions
・Teaching Materials

The common point among these is that accurate information transmission is prioritized, and there is no need for a "good" way of reading or a way of reading that conveys emotions.
In recent years, AI automatic voice tools can create fluent voices that sometimes make you think "better than an amateur can read". Therefore, for using such delicate appeal in things that do not require it, there are hardly any quality deficiencies.

Now, let us introduce the advantages of AI automatic voice in specific points.

Benefits

There are several benefits to using AI automatic voice instead of human narration for voice creation in multilingual localization projects:

・Time and Cost
・Ease of revision, standardization of quality

■Time and Cost

Time and cost are easy to imagine, aren't they?

Narration recording by a person requires a lot of effort and time, from preparing the recording equipment/environment to actually recording the narration. There may also be a need to adjust schedules to be present during the recording.
On the other hand, when using an AI automatic voice tool, all it takes is inputting the text you want to convert into speech and pressing the voice generation button, and the voice will be completed in just a few seconds. The time and cost benefits of this are extremely significant.

■Ease of modification, standardization of quality

As the second benefit, we mentioned ease of modification.

When checking the created voice once, there may be requests such as "I want to change the pronunciation of this word", "I want to change it to a more friendly expression", and "I want this sentence to be read more slowly". However, reflecting those modifications in the re-recording may not be easy for narration by humans. This is because in order to re-record, it is necessary to secure the recording studio and narrator's schedule again. Even for a single change, in multilingual localization that expands to many languages such as dozens, it often becomes unrealistic to make modifications as it would require redoing everything in all languages.

In contrast, with AI automatic voice tools, you can modify the text read by the machine, input speed instructions, and generate the voice again, reflecting the modifications immediately. This is a significant advantage in multilingual projects that tend to require many modifications.
Furthermore, if you have concerns about differences in quality depending on the language with narration done by humans, using AI automatic voice tools allows for fine adjustments, enabling the standardization of quality across languages.

3. Preparations for AI Automatic Voice Utilization

"I understand the benefits and use cases of AI automatic voice tools. I want to try using it right away, but how should I proceed?" I would like to answer such questions.
Although it is a very convenient AI automatic voice tool, it is possible to aim for further efficiency and improvement in the quality of the final output through pre-preparation.

Schedule

We introduced the "ease of correction" as a major benefit of AI automatic voice. Adjustments to fine details that were difficult due to scheduling constraints with human narration can now be made possible by utilizing AI automatic voice tools... There is no reason not to take advantage of this benefit.
AI automatic voice tools, depending on the product, are often based on cost systems determined by volume or duration. (The differences in cost systems depending on the product will be explained in the next blog post)
Therefore, it is necessary to create a schedule assuming that there will be modifications after voice generation, and to select and purchase AI automatic voice tools accordingly. It may require some extra effort, but compared to projects with human narration, there may be overall time and cost benefits.

Confirming the pronunciation of abbreviations

Even for commonly known abbreviations, the pronunciation may vary depending on the language.
One example is "ISO". As shown in the table below, English and Indonesian commonly use different pronunciations compared to Japanese.

Japanese	English	Indonesian
ISO	ISO/ ISO	Iso/ Aisuo

 

In cases like this, if you output the pronunciation without specifying which pronunciation to unify in advance, there is no guarantee of which pronunciation will be output or whether the pronunciation will be unified throughout the entire audio. Therefore, it is important to register/specify the pronunciation in advance on the tool side to prevent unnecessary modifications later on.

Checklist

In projects that use AI automatic voice, just like projects with human narration, it is essential to confirm the created voice.
Some people may have the impression that "since it is being read mechanically, there should be no reading mistakes." However, in reality, there are still points that automatic voice cannot handle, such as differences in intonation depending on the context, unnatural gaps or lack of necessary gaps that make it difficult to understand.
Especially in multilingual localization, there are often cases where the meaning changes due to differences in intonation depending on the language, so checking after generation is indispensable.
However, if the person in charge of the check randomly checks, there is a risk that unnecessary modifications will have to be made based on the person's sense, or conversely, they may be overlooked with the thought of "this level is acceptable." To ensure uniform quality across multiple languages, it is very important to align the check points between languages.
By creating a checklist that lists the necessary check points, you can smoothly proceed with the creation of multilingual voices. We will introduce specifically what kind of check points are necessary in the next blog.