📂 Directory

• NOTICE
• ABOUT
• DOWNLOAD
• INSTALLATION
• CONFIGURATION
  • Required
  • Optional
• USAGE (CPU MODE)
  • Local Mode
  • Local Webtoon Mode
  • Web Mode
    • Real-Time Translation
• USAGE (GPU MODE)
• UPDATE
• EXTRA INFO
  • How to Get Gemini API Key
  • Webtoon Mode

NOTICE

Some things have been changed & fixed, so it's recommended to update to newer components if you have installed it before. See the UPDATE section for more info.

Since the guide has become too long and complex than originally intended, I decided to simplify it. You can see the more detailed version of this guide here.

ABOUT

This fork doesn't change the core functions of the original program. This is still Manga Image Translator, but with some minor tweaks & extra components to make it easier and more convenient to set up and use.

Note

For more info about the main usage of MIT, always refer to https://github.com/zyddnys/manga-image-translator.

The Changes:

• Add installer
• Add updater
• Add launchers
• Add .env file
• Add PyTorch checker
• Add folder selection feature
• Add XPU (Intel GPU) support (untested)
• Improve handling of webtoon format (🛠️working but need improvement)
• Sort input folders in natural order
• Use recommended configurations by default
• Disable some functions in order to bypass errors
• Clean up result folder except for log file by default

Important

The installer only supports Windows 10 & 11.

DOWNLOAD

1. Click on the green button on the top.
2. Select "Download ZIP".
3. Right click on the downloaded .zip file.
4. Select "Extract Here" with WinRAR or 7-Zip.

INSTALLATION

1. Open PowerShell as Administrator.
2. Change PowerShell execution policy by entering the command below:

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope LocalMachine

3. Enter y or yes.
4. Close the PowerShell.
5. Right click on MIT-installer.ps1.
6. Select "Run with PowerShell".
7. Select "Yes" if UAC prompt pops up.
8. Wait until you get ${{\color{lightgreen}{\textsf{INSTALLATION COMPLETED!}}}}$ message.

Tip

If you get a warning when opening the installer, uncheck the option, then Open. If you don't do this, the script won't be able to run properly.

View image

CONFIGURATION

Required

1. Open .env file with text/code editor (Notepad, VS Code, etc).
2. Paste your Gemini API key between the quotation marks.
3. Save.

Optional

1. Open my-config.json and gpt_config-example.yaml in examples folder & settings.json in my_tools folder with text/code editor.
2. Change the settings as you see fit.
3. Save.

USAGE (CPU MODE)

Note

• All local modes support batch translation.
• The first time you run the program, it will automatically download the selected detection, OCR, & inpainting models. After that, it won't need to do it again, unless you have changed the relevant configurations in my-config.json.

Local Mode

1. Right click on MIT-local-launcher.ps1.
2. Select "Run with PowerShell".
3. Select a folder containing your manga/hwa/hua.

Local Webtoon Mode

Warning

This launcher has a really high RAM usage!

1. Right click on MIT-local-webtoon-launcher.ps1.
2. Select "Run with PowerShell".
3. Select a folder containing your manga/hwa/hua.

Web Mode

1. Right click on MIT-web-launcher.ps1.

2. Select "Run with PowerShell".

3. Visit http://127.0.0.1:8000 (default).

4. Press Q to stop the server.

   Real-Time Translation

   There are a bunch of available browser extensions out there, but I will use ComicReadScript here. In general, they have similar configurations.

   • Install Tampermonkey from https://www.tampermonkey.net

     Next, if you use Chromium-based browser, then you need to follow the steps here: https://www.tampermonkey.net/faq.php#Q209.

   • Install ComicRead from https://sleazyfork.org/en/scripts/374903-comicread
   • Visit RAW manga/hwa/hua website.
   • Select a chapter from any available series.
   • Click on Extensions menu bar > Tampermonkey.
   • Select "Enter simple reading mode"
   • Hover your mouse over the left side of the page.
   • Click on "Scroll mode" button.
   • Hover over the left > "Settings":
     View config

     

   • Click on "Settings" button to close.
   • Click on "Translate current page" or "Translate current page to the end".

Tip

• If your PC is slow, it won't actually be real time. So, consider using GPU if you have a powerful one.
• Check the PowerShell to see more detailed progress of the translation process.

USAGE (GPU MODE)

Go to here.

UPDATE

Warning

This updater will replace the old files with the newer ones, so make sure to back up the files you want to keep first. For more info, see here.

1. Download MIT-updater.ps1.

   only if there is no ### VERSION ### in your MIT-updater.ps1.

2. Move it to your manga-image-translator-main folder.
3. Right click on it > Run with PowerShell.
4. Wait until you get ${{\color{lightgreen}{\textsf{UPDATE COMPLETED!}}}}$ message.

EXTRA INFO

How to Get Gemini API Key

1. Visit https://aistudio.google.com/app/apikey.
2. Accept the Terms and Conditions.
3. Click "Create API key".
4. Name your key.
5. Choose project > Create project.
6. Select the newly created project.
7. Click "Create key".
8. Click the code in the "Key" column.
9. Click "Copy key".

Tip

Gemini API Free Tier has rate limits, see: https://ai.google.dev/gemini-api/docs/rate-limits#current-rate-limits.

To check your quota:

1. Visit https://aistudio.google.com/app/usage
2. Make sure you are on the right account & project.
3. Click "Open in Cloud Console" on the bottom.
4. Scroll down > Click "Quotas & System Limits".
5. Scroll down > You will see your model quota usage on the top result. If you don't see it, use Filter to search it.

For example:

View image

Webtoon Mode

Warning

This mode will attempt to merge all images in each chapter folder into one really long image respectively first. MIT then will have to load and process the long-ass images for translation, which inevitably causes it to consume a lot more RAM and time than regular mode. Last but not least, it will merge the translated images into one* before splitting all translated images back into the same number of parts as the original images in each folder (the height and the split position won't be identical tho). For more info, see here.

_{*if the specified merged image number is greater than 1.}

Pros

• Better translation result because the translator will get all texts from one chapter at once, so it will have more contexts than when it receives the texts from only one page at a time.
• Better OCR result in a way as there is no potentially split speech bubble resulting in incomplete text detection.

Cons

• Slower and heavier.
• Speech bubbles are dirtier.
• Some texts are not detected and/or inpainted at all just like in regular mode. It's just that the missed areas will be different because the difference in their image heights. That's why it's recommended to increase or decrease detection_size & inpainting_size in my-config.json to improve the result. See https://github.com/zyddnys/manga-image-translator?tab=readme-ov-file#tips-to-improve-translation-quality.
• Prone to server overloaded error. (just retry it XD)
  It seems that it's not really caused by the launcher, or is it? 🤔, since even the paid users are experiencing the same issue, see: google-gemini/gemini-cli#4360. Alternatively, you can change the model in .env file or the translator in my-config.json.

Note

The webtoon mode can use up to around 20GB RAM on my laptop.

My PC Specs:

• Model: ASUS VIVOBOOK 14X M1403QA
• CPU: AMD Ryzen™ 5 5600H (6C/12T)
• GPU: 512MB AMD Radeon™ Vega 7 Graphics (integrated)
• RAM: 24GB DDR4 3200 MT/s
• Storage: 512GB M.2 NVMe™ PCIe® 3.0 SSD
• OS: Windows 11 Home Single Language 64-bit (25H2)

If your PC specs are equal or better than mine, then you should be fine, probably 😅.