BasicCAT — Computer-Aided Translation (CAT) Tools

Browser Extension for Manga Translation

Sat, 16 May 2026 12:22:50 +0000

There are many manga that never get translated. When we browse manga websites through a browser, we wish to see translated versions of manga images directly, but browsers don’t have this feature built in. That’s where browser extensions (add-ons) come in.

The computer‑aided image translation software ImageTrans provides such a browser extension, supporting Chrome, Edge, Firefox, and other browsers.

Manga translation typically involves the following steps:

OCR text recognition
Layout analysis – merging text based on semantics and adjusting text order
Translation (using large language models such as ChatGPT, Deepseek, etc.)
Text removal (inpainting)
Back‑filling the translated text (lettering)

Depending on the language and purpose, we can use different software combinations and translation strategies.

For example, the text removal step can be time‑consuming if processed with AI. If our goal is merely to understand the meaning, we can use a white background fill instead. The same applies to OCR and translation.

For OCR, if the language is a mainstream one like Chinese or English, local OCR such as PaddleOCR works quickly with good recognition accuracy. For vertical Japanese, traditional Chinese, etc., higher‑accuracy models are usually needed – for instance, mangaOCR for Japanese.

For translation, ordinary machine translation often performs well for languages with strong logical structures like English. For high‑context languages like Japanese, especially when OCR may contain errors, a high‑quality large language model is required – sometimes even feeding the image to the model for better semantic understanding.

The browser extension of ImageTrans supports freely combining various OCR and translation technologies to meet the translation needs of various manga images. The entire process can run on your local computer, free to use, no subscription required, no pay‑per‑page fees, with faster response times and better controllability. This is the method I recommend for translating images in web pages.

Below is a demo video:

The video demonstrates how to use ImageTrans’s browser extension to translate manga images from Pixiv, from Japanese to English. It shows purely browser‑based local processing methods – region‑based screenshot translation, full‑image translation – as well as using the ImageTrans desktop edition for higher‑quality translation.

Remove Text in Manga and Preserve Screentone

Mon, 11 May 2026 11:21:50 +0000

In the lettering step of manga translation, the original text needs to be erased and replaced with the translated text.

One of its tedious tasks is restoring screentones. Most tutorials use Photoshop’s clone stamp tool, which requires careful alignment with the screentone and struggles with complex patterns like gradients. Photoshop’s built-in content-aware fill is also unpredictable and may fill in irrelevant image content.

Another method involves using custom patterns and overlaying them on a separate layer, as explained in the Krita documentation, but the supported patterns are limited. More professional software like CLIP STUDIO PAINT and ComicWorks also require significant manual effort.

With advances in image processing and AI technology, several software tools now offer effective screentone restoration.

The computer-assisted manga translation software ImageTrans provides an all-in-one manga translation platform, covering processes from scanning, text recognition, translation to typesetting.

It integrates several methods for text masking and text removal. Below is a comparison of the results produced by some of its methods that handle screentones well.

Original image:

Text mask:

PatchMatch (traditional image processing):

Flux Klein 9B：

Lama Inpaint (original):

Lama Inpaint (fine-tuned for Japanese manga):

Gemini 3.0 (Nano Banana):

As you can see, the results are quite good, especially the fine-tuned Lama’s, and even when there are minor flaws, only a small amount of correction is needed.

These generative AI models have powerful repainting capabilities. Besides restoring screentones, they can also reconstruct various backgrounds covered by text, such as human figures, landscapes, and other content.

How to Translate Images on a Web Page

Fri, 08 May 2026 12:42:50 +0000

Browsers all have built-in webpage translation features, but they usually can only translate selectable text, not images. Yandex supports image translation, but with unsatisfactory results. Safari allows copying text but has no translation function. The computer-assisted image translation software ImageTrans provides a Chrome browser extension that supports direct translation of images on web pages, erasing the original text, filling in the translated text, and restoring colors, rotations, and various styles.

Below is its processing workflow:

Through the browser extension, an image translation request is sent to the server.
The server calls the connected ImageTrans program to perform the translation.
ImageTrans carries out text recognition with OCR, text translation, text inpainting and typesetting, generating a translated image.
After receiving the translation result, the browser extension replaces the original image with the translated version.

Because this operates within the browser, some technical limitations must be overcome:

Browsers restrict the use of local programs. The local server needs to add headers that allow local invocation. Older versions of Chrome also require HTTPS.
If an image is cross‑origin restricted, it cannot be downloaded directly, and fetching image data via canvas will result in a tainted canvas error. In this case, the extension must modify headers to remove CORS restrictions. If referrer information is also restricted, or if the image is not an <img> element but a canvas, making direct download impossible, a screenshot feature can be used—but content outside the viewport will be lost. Pixiv, for example, restricts cross‑origin access.
Some websites overlay additional content on images, so replacing the image alone does not display the translated version. For instance, X.com.

The ImageTrans extension can already translate images on common websites. Tested sites include Read Comics Online, Pixiv, MangaDex, Lezhin, Tencent Comics, and more. It supports various image types: manga, webtoon, manhua, comics, illustration, scanned books, etc.

Many browser extensions on the market rely on remote servers and require a subscription. ImageTrans, by contrast, runs entirely on the local computer, allowing unlimited translation, using various OCR engines and large language models, directly downloading images to the local machine, preserving translation results, and freely defining text styles.

Extension link: https://github.com/xulihang/ImageTrans_chrome_extension

Demo video:

PS:

ImageTrans’s extension also supports OCR right in the browser through WebAssembly. But its pocessing speed is slow and the ability is limited. It is provided as a free option along with the public ImageTrans server. It is recommended to use the extension with ImageTrans desktop version.

Real-Time Screen Translator

Sat, 11 Apr 2026 12:39:50 +0000

Recently, I want to play the raw Japanese GBA game Ace Attorney. I like the feeling of the original Japanese version and want to improve my Japanese skill. My Japanese level is limited so I still need a screen translator to help me.

ImageTrans is such a program. It integrates with various OCR and translation engines and can work as a real-time screen translator to play games like visual novel and read manga.

Here is its workflow:

The program detects whether the screen capture of the region specified by the user has changed and is now stable.
Use OCR to recognize the text in the region.
Use AI translation engines like ChatGPT and Deepseek to translate the text.
(Optional) Add pronuciation annotation (Furigana for Japanese or Pinyin for Chinese) and use text-to-speech engine to read the text.
(Optional) Save the recognition results for learning afterwards.

The whole process is automatic. You can also make the screen capture by yourself and ask the program to translate, since you may also need to read the text outside the region you specified.

The following is a demo video. Although it takes a while for the OCR and translation to complete and there are mistakes, it is playable.

ImageTrans can also translate other medias, like books and manga. It is an integrated translation tool, like Photoshop for translation.

ImageTrans can run on desktop platforms like Windows, macOS and Linux. Mobile operating systems like Android and iOS are currently not supported.

Optimizing Translation Speed and Quality

Real-time translation processes text only within the current input box, lacking contextual information, which leads to lower translation quality. Additionally, performing a translation operation each time results in slower speeds.

For many games, the text can be extracted, or transcriptions are already available. In such cases, we can pre-translate the text using a large language model. Later, during on-screen translation, the system can match the source text with OCR results and directly retrieve the pre-translated text, achieving both high speed and high quality. This feature is called translation memory in ImageTrans.

Below is a simple comparison using DeepSeek for translation. It shows that more context leads to higher translation quality.

Original Text	Full-Text Translation	Sentence-by-Sentence Translation
（うう‥‥。キンチョーするなあ‥‥）	(Ugh… I’m so nervous…)	(Ugh… Don’t be so nervous…)
‥‥なるほどくん！	…Naruhodo-kun!	…Mr. I See!
ふう。なんとか、間にあったわね。	Phew. I managed to make it in time.	Phew. We made it just in time.
どうかしら？　初めての法廷は。	How do you feel? Your first time in court.	How about it? Your first time in court?
こ、こんなにドキドキするの、小学校の学級裁判のとき以来です。	Th-This is the first time my heart has raced like this since the class trial in elementary school.	My heart’s pounding like this for the first time since the class trial back in elementary school.
‥‥それはそれは。ずいぶん、ごぶさたしてるのね。	…Well, well. It’s been a long time, hasn’t it?	…My, my. It’s been quite a while, hasn’t it?

Additionally, ImageTrans supports fuzzy matching between the source text and OCR results. Even if OCR makes errors, it can still find matches, preventing translation mistakes caused by faulty OCR.

For example, the sentence 借りがあるんですよ。ヤツには。 might be misrecognized as 借りがあるんですよ。ャツには。, where the character “ヤ” becomes the small kana “ャ”, potentially leading to translation errors. For instance, DeepSeek might mistakenly translate it as “I owe you a favor” instead of “I owe him a favor.” Matching against pre-translated text extracted from the original avoids such issues.

Manga Typesetting Tool Comparison: TypeR and ImageTrans

Tue, 07 Apr 2026 13:06:50 +0000

In the world of manga translation, typesetting is where everything comes together. After OCR and translation, you still need to place text back into the artwork in a way that feels natural, readable, and faithful to the original design. This is where tools like TypeR and ImageTrans come in.

Both target similar users—scanlators, translators, letterers, and comic localizers—but they represent two very different philosophies: one centered on Photoshop, the other on an integrated pipeline.

Overview

TypeR is a Photoshop extension (forked from TyperTools) designed specifically for manga and comic typesetting. It enhances Photoshop with features like smart alignment, style management, and multi-bubble handling, aiming to speed up manual work without leaving the design environment.

ImageTrans is a standalone computer-aided translation tool that covers OCR, translation, and text reinsertion. It can automatically detect text regions, remove original text, and inject translations, while still supporting export back to Photoshop for refinement.

Workflow Comparison

TypeR: Photoshop-Centric Workflow

TypeR assumes Photoshop is your main workspace:

Open image in Photoshop
Insert text using TypeR tools
Adjust layout manually

Its biggest strength is eliminating repetitive actions (copy-paste, alignment, style switching) inside Photoshop.

However:

OCR and translation must be done elsewhere
Workflow is fragmented across tools

ImageTrans: End-to-End Workflow

ImageTrans covers the full pipeline:

Load images
Detect and extract text (OCR)
Translate using multiple engines
Remove original text and reinsert translation
Export to PSD for final refinement

It integrates OCR, large language model translation, text cleaning, typesetting and batch processing into one system.

This reduces tool switching and enables large-scale workflows.

Feature Comparison Table

Feature	TypeR	ImageTrans
Core Type	Photoshop plugin	Standalone application
OCR (text extraction)	❌ Not supported	✅ Built-in multi-engine OCR
Translation integration	❌ External only	✅ Multiple MT engines + CAT features
Text detection	❌ Manual	✅ Automatic detection (AI + heuristics)
Text removal (cleaning)	❌ Manual in Photoshop	✅ Automated (masking, lama inpainting, nano banana, etc.)
Text alignment	✅ Yes	✅ Yes
Text reinsertion	✅ Manual typesetting	✅ Automatic + adjustable reinjection
Multi-bubble handling	✅ Yes	✅ Yes
Style management	✅ Advanced (folders, presets)	✅ Global styles
Batch processing	❌ Not designed for it	✅ Supported
Photoshop integration	✅ Native plugin	✅ Export to PSD
Automation level	Low (manual control)	High (semi/fully automated)
Cross-platform	Depends on Photoshop	✅ Windows / macOS / Linux

Which One to Use

Choose TypeR is you prefer to do the typesetting entirely in Photoshop.

ImageTrans is a better choice in almost all aspects:

An integrated environment for document scanning, image extracting from CBZ/PDF, image editing, OCR, translation, cleaning and typesetting.
Compatible with basiccaly all Photoshop versions (CS2+) and any other image editors like Krita and GIMP.

How to Convert Vertical Japanese PDFs to Horizontal Layout

Wed, 25 Mar 2026 06:25:50 +0000

Japanese is typically written in vertical, which may be difficult for people accustomed to horizontal text. We can convert it to a horizontal layout.

PDFs can be divided into two types: those with selectable text and those that cannot be copied. For the selectable type, since the PDF simply renders text at fixed positions, the text often cannot be copied in reading order and requires layout analysis. For non-selectable versions, such as scanned PDFs, OCR software is needed to recognize the text, followed by layout analysis to determine the correct text order.

The computer-aided image translation software ImageTrans supports processing both types of PDFs, exporting vertically oriented text into a horizontal format.

Example:

Original PDF File:

Converted horizontal text:

もちろんです以下にランダムな日本語の文章を生成しました。

一匹の子犬。王都が見えてきた。でかいのは態度だけで後は何もかもが小さい。

It also supports adding furigana annotations:

もちろんです以下いかにランダムな日本語にほんごの文章ぶんしょうを生成せいせいしました。

一匹いっぴきの子犬こいぬ。王都おうとが見みえてきた。

でかいのは態度たいどだけで後あとは何なにもかもが小ちいさい。

How to Add Pronunciation (Furigana) to Japanese Kanji

Wed, 25 Mar 2026 06:25:50 +0000

Japanese kanji originated from China and are logographic characters. Their pronunciation can be annotated using romaji or kana. The kana used for this purpose are called furigana.

Below are some examples of text with pronunciation annotations (raw text and annotated):

Example 1:

一匹の子犬。

一匹いっぴきの子犬こいぬ。

Example 2:

もちろんです以下にランダムな日本語の文章を生成しました。

もちろんです以下いかにランダムな日本語にほんごの文章ぶんしょうを生成せいせいしました。

Example 3:

王都が見えてきた。

王都おうとが見みえてきた。

Example 4:

でかいのは態度だけで後は何もかもが小さい。

でかいのは態度たいどだけで後あとは何なにもかもが小ちいさい。

Adding pronunciation annotations to Japanese kanji is not a simple task. It requires correctly segmenting Japanese sentences into words and then obtaining the accurate pronunciation of the segmented results. Japanese kanji pronunciation is complex, with on’yomi (Chinese readings) and kun’yomi (Japanese readings), and phonetic changes often occur when multiple characters are combined.

Kuromoji is a specialized Japanese morphological analysis software that performs word segmentation, grammatical analysis, and pronunciation retrieval based on dictionaries such as ipadic and unidic. It can be used to annotate kanji with pronunciation.

However, it has some drawbacks.

The default ipadic dictionary has limited data, leading to inaccurate annotations. Unidic, on the other hand, segments too finely—for example, splitting “日本語” (Japanese) into “日本” and “語,” which results in annotation errors. Additionally, it handles counters like “一匹” (one small animal) poorly.

ImageTrans integrates Kuromoji and combines the results from ipadic and unidic to generate accurate annotations. It also automatically handles counters, achieving better annotation results. This can be used to assist with manga reading, Japanese language learning, and automatic text annotation. If higher accuracy is required, large language models such as GPT or DeepSeek can also be used for annotation.

How to Detect Text Colors in Images

Sun, 15 Mar 2026 08:28:50 +0000

Detecting text colors in images, including background colors and stroke colors, can be used for layout restoration, translation, and other operations.

Original image:

Translated image:

Implementation Principle

For background colors, KMeans can be used to cluster colors and identify the dominant colors in the image.

For text colors, contour detection can be performed to calculate the average pixel value of the text contour and exclude pixels that are similar to the background color.

However, if the background is complex and stroke colors or the color of each character need to be extracted, traditional image processing methods may not perform well. In such cases, convolutional neural networks can be used for extraction.

The computer-assisted image translation software ImageTrans integrates the algorithms mentioned above.

Examples

Below are some operational examples.

Example 1

In many CG images, different colors are often used to distinguish dialogues between different characters.

ImageTrans can detect stroke colors and text colors. Below are the raw recognition results.

As can be seen, the text colors and stroke colors have been detected, but they are not very accurate.

In the project settings, we can predefine several styles and specify the colors used for each style.

Afterward, perform a color matching operation to match the styles based on the text colors.

This resolves the issue of inaccuracies in the detected colors.

Example 2

Some texts use rich text formatting, where a single line of text may contain multiple colors. ImageTrans supports recognizing the style of each character and outputting results with rich text tags.

How to Convert Traditional Chinese PDF from Vertical to Horizontal and Enhance Text Clarity

Sun, 08 Mar 2026 11:32:50 +0000

I recently came across several older Taiwanese academic papers and found that they were all in PDF format with vertical text, Traditional Chinese, and black-and-white scanned copies. The text was illegible, which was so difficult to read that I worried it might worsen my myopia. So, I came up with some ways to optimize the reading experience. For users in Mainland China, converting the vertical text to horizontal format makes it much easier to read. However, sometimes preserving the original layout is better. So I also tried using OCR to recognize the text, remove the original content, and reformat it with clear, readable text. Below is a demonstration of the results and how to do it.

Demonstration of Results

Original file (PDF):

Converted to horizontal markdown format:

卷五十第

期一十二第

## 那一個甘蔗品種好?

## 選擇蔗種的基本原則

選擇甘蔗品種的基本原則,是「適地適種」,而「適地」的意義,並不是單指適合的土壤的意思,還包括蔗園所在地點的一切環境條件,諸如氣候狀況、灌溉條件、肥料多少、勞力充份否,間作不間作,留宿根不留等等,都要考慮到.只有適合種蔗環境的品種,才能有好的收成,所以在選擇品種之前,首先要明白所經營的蔗園的環境.下面幾點是主要的考慮對象：

（1）土質和地力：土質的粘重或輕鬆,主要關係土壤水分的保持與喪失.粘重的土壤,保水力较强,纵然在灌溉水不够充分的地方,也能栽培比較不耐旱的品種,但是粘重的蔗園容易排水不良,根系太深,不耐浸的品种就发育不良.土质轻松的土壤適恰相反,不適合栽培不耐旱品種,而適應根系深的品種.

甘蔗品種的耐肥性是個個不同的,有的適合在好地方的蔗園,有的在地方好的蔗園反而減產.所以地方的肥沃瘦瘠也需要密切注意.

（2）灌溉的條件：要說甘蔗是水做的,一點也沒有錯,一支甘蔗有七〇％以上是水.因之,水成了種蔗的必具條件,只是甘蔗品種裏有些特別喜愛水一些,缺了水,產量隨即降低,另一些對水淡漠一些,早一點影響產量不大.選擇品種的時候,不能不顧到這點.

（3）宿根：每一個品種的宿根能力多有不同,有些可以宿根,有些宿根产量很低,留宿根化不來,要是選錯了就要自己吃虧.

（4）植期不同：有些品种只能作为秋植,而且必須早秋植；延遲種植,產量降低.另一些品種又只適合春植,作為秋植時並不理想.而二期糊仔有二期糊仔的品種,一期糊仔有一期糊仔的品種,

都混雜不得的.

（5）勞力的多少：有的蔗友家中勞力很多,或者財力雄厚,足以招雇够量的勞力來從事精耕栽培,那麼就應該選擇適合精耕栽培的品種.否則,應該選擇在粗放耕作時仍有相當產量的品種.

（6）病蟲害問題：現在推廣的品種,大致說對於主要的病蟲害都是抵抗的,但在容易發生病蟲害的環境如玉米栽培地區,有些品種比較容易發生露菌病,應該避免栽培.

## 目前推廣的優良蔗種

談完了選擇品種時考慮的主要種蔗環境條件,現在來選擇品種,首先要介紹一些最近由糖業試驗所育成,在推廣中的甘蔗新品種.

F一四六已經推廣三個年期了,大家對它都已十分認識,它是一個中大莖,多分櫱的品種.初期生育緩慢,而在次年高溫多水的季節迅F一四六速生長.原料莖很多,只是短一點,適合在地力中等以上,有水灌溉,不很粘重的地方作为早秋植,F一四六可以間作,宿根也不錯,最高的產量可以達到三十萬公斤.

F一四八作為春植或晚秋植時,它增產能力比在早秋植時高,同時也不適合二期糊仔.蔗莖又細小一點,所以蔗友們不很歡迎.事實F一四八上,用F一四八作晚植蔗園的品種,它的產量比F一四六,和F一五二安定.而且在所有的新品種中,F一四八是糖份最高的一個,也是最不怕露菌病的品種,又因為它很早熟,種了它,多半是在製糖季初收穫,蔗友們可以提前利用種蔗後的土地.

F一四八也不適合在粘重或過份輕鬆,以及乾旱的地方栽培,間作和宿根的條件不及F一四六.

## F一五一

F一五一適合在粘土蔗園秋植及二期糊仔

,在溫度較低,土壤較乾的時候,萌

芽比其他新品種好.它是中莖種,原料莖比F一四六細而長,也比F一四六容易倒伏,但不易开花；和F一四六同样地晚熟,步留也差不多.露菌病的感染程度較重.F一五一的新植產量在輕鬆的壤土裏不及F一四六,在粘重的土壤則勝過F一四六.宿根的產量和F一四六相仿.

F一五一不能作為晚植蔗園的品種.

F一五二是地力肥沃,灌溉水充沛蕉園的增產王牌,它的原料莖高而粗,只要保持適當的原料莖數,很容易的得上二十萬公斤以上的F一五一甲當產量.不過在灌溉水缺乏,地力不夠而又沒有辦法施用重肥的蔗園,F一五二沒有增產的希望.

七、八月種植的F一五二容易母莖徒長,因此不如九月種植的蔗園產量穩定.F一五二是個好

## ·紹介者作期本·

▼湯冠雄：曾任臺灣糖業試驗所種藝系技師,現任該所虎尾蔗作改良場場長,從事甘蔗栽培研究已十七年.

![图片]()

夏輔禹：安東省岡城縣人,臺灣省立農學院畢業,現任糖業試驗所虎尾蔗作改良場研究組組長,一直擔任蔗田間作試驗方面工作.

![图片]()

A clear version preserving the original layout (Image + PDF):

Text-based PDF with vector text, remains sharp even when zoomed in:

File Link

How-to Guide

This requires using the OCR software ImageTrans.

Use the built-in bubble detection feature to detect vertical text lines.
Use a large language model for OCR. Since the text here is quite blurry, only large language models can achieve good results. I chose the qwen3-vl-235b-a22b-instruct model for this task.
Use PaddleOCR’s DocLayout panel detection model to identify text paragraphs and sort the text lines from right to left according to the reading order.
Specify text styles for both vertical and horizontal text. Then, determine which text is horizontal and which is vertical based on the aspect ratio.
Export as PDF.
Next, export as markdown. Here, we can merge text lines by paragraph to remove unnecessary line breaks.

Software main interface:

Displaying panels and order:

How to Scan Negatives

Fri, 27 Feb 2026 12:28:50 +0000

Photos taken with old film cameras usually leave behind negatives (film) after being developed and printed at a photo lab. I recently came across quite a few of these negatives. The negatives contain a type of silver salt that is highly light-sensitive. Standard negatives are 35mm wide and 24mm high, and the developed content is inverted.

Because negatives are very small and have an additional orange mask (color cast), digitizing them is not an easy task. A common method is to use a professional film scanner. If a scanner isn’t available, you can also use a mobile phone or a camera to “scan” them by photographing.

Photographing with a Camera

Negatives are transparent and require transmitted light to reveal the image. You can prepare a white light source, such as a phone screen or a ceiling light.
Use a macro lens to take the photo. For a mobile phone, you can attach an external macro lens.
After shooting, use software like Adobe Photoshop for further processing: inverting the colors and using the “Auto Color” function to remove the orange mask.

Since a mobile phone was used here, there might be issues like insufficient sharpness, inaccurate colors, and screen moiré patterns. Using a professional camera would yield much better results, potentially offering greater sharpness than some scanners. Additionally, professional cameras can save files in raw format, which is more convenient for color correction.

Scanning with a Scanner

Choose a flatbed scanner that supports scanning negatives, such as the Epson V850. The one I used here is an Epson V300, which comes with a transparency unit and film holders, making it suitable for scanning negatives.

You can use the manufacturer’s software, like Epson Scan, to scan. It can automatically crop negatives, invert colors, and remove the orange mask.

If you prefer to obtain only the raw scan data, you can use software like ImageTrans to control the scanner via interfaces such as TWAIN, ICA, or SANE.

Afterwards, you can use ImageTrans or other software for cropping, inverting, and color correction.

Scanning using ImageTrans:

Result after cropping, inverting, and color correction using ImageTrans: