This article is translated by AI.

This article is based on the previous five articles and introduces methods to improve translation speed and quality through some auxiliary tools.

Archive Directory

Japanese Manga Machine Translation Experience (0) - Table of Contents

Translation Separation

In the previous articles, I have been emphasizing the function of exporting original text and importing translation of these four tools. In this article, we will know the importance of this function in translation separation.

Traditional integrated translation has several disadvantages. First, there is very little context (except for manga-image-translator, which can manually set the number of contexts), or even no context, such as Saber-Translator; then the glossary support is poor, Saber-Translator and BallonsTranslator do not have the function of glossary, which is easy to cause problems such as confusion in the translation of names; finally, the translation speed is slow, these four software are single-threaded translation, translating only one sentence or one picture at a time, which is slower than multi-threaded translation.

Therefore, we need to separate the original text to be translated, use LinguaGacha and KeywordGacha these two auxiliary tools to translate separately, and then import the translation.

The installation of LinguaGacha and KeywordGacha is not difficult, so I will omit it here.

For the convenience of demonstration, I use a chapter of 12 pictures as the manga to be translated, and BallonsTranslator as the translation tool. In fact, the larger the number of pictures, the more obvious the effect of this method. In addition, in the table in Japanese Manga Machine Translation Experience (1) - Overview, the translation tool ImageTrans supports the function of exporting original text and importing translation best, but because it is paid, it is not convenient for readers to experiment, so I chose the free BallonsTranslator as the demonstration tool.

Step 1: In "Run", only check "Enable Text Detection" and "Enable OCR".

Step 2: Run, start running the program.

Step 3: After the operation is finished, export the original text as Markdown.

Step 4: Use KeywordGacha to get the glossary, the result is as follows.

src	dst	info	regex	count
一瀬	Ichinose	Family		6
小海	Koumi	Unknown gender name		1
小湊	Kominato	Family		2
結月	Yuzuki	Female name		9

Step 5: Translate the original text using the glossary in LinguaGacha.

Step 6: Import the translation.

Step 7: Enable inpainting.

This is the end, below is a translation result.

In the actual use process, it may be necessary to formulate corresponding text protection rules or convert Markdown to json for formatted output.

Contributors: Ming Wei

Previous pageManga Reading and Management Experience

Next pageJapanese Manga Machine Translation Experience (5) - ImageTrans