Kamite Save

Japanese immersion assistant for learners (Windows/Linux)

Project README

Kamite

A desktop language immersion companion for learners of Japanese

[!NOTE] This is alpha software.

Kamite is desktop software to aid learning Japanese through immersion in native media. It brings Japanese text from those media into a web browser interface, enabling lookup with pop-up dictionaries (such as Yomichan) and websites (such as DeepL Translate), and more.

https://user-images.githubusercontent.com/4130634/178029301-075cb207-a154-42d2-adb5-ce8fdbcd722f.mp4

(Featured in the above demo video: [All shots] Firefox, Sway. [Shot 1] Gomics-v manga reader (Kaguya-sama wa Kokurasetai, ch. 140); Yomichan (dictionaries: JMDict, Kanjium Pitch Accents, Narou Freq, VN Freq). [Shot 2] mpv (Suzumiya Haruhi no Yuuutsu (2006), ep. 9). [Shot 3] Summer Pockets; Textractor; contrib/kamite-toggle-visibility.sh script; waycorner.)

Highlights

Kamite supports Linux* and Windows (macOS support planned for the beta release).

(*Linux: Xorg, wlroots, GNOME Wayland, Plasma Wayland. OCR-ing arbitrary screen areas not supported on GNOME Wayland and Plasma Wayland.)

Kamite is cost-free and is licensed under the GNU AGPL v3 or later.

Table of contents

  1. Installing Kamite
  2. Updating Kamite
  3. Launching Kamite
  4. Troubleshooting
  5. User interface overview
  6. Text extraction
  7. Text use
  8. Custom commands (Launching external executables)
  9. Keyboard shortcuts
  10. Launch options
  11. Config
  12. Style customization
  13. Command API
  14. Privacy
  15. Development
  16. License

Installing Kamite

Linux

Arch User Repository

An AUR package is available under the name kamite-bin (installs to /opt/kamite).

Debian/Ubuntu-based

A Debian (.deb) package is available among the files on the Releases page. Install with:

sudo apt install /path/to/[package-file].deb

Uninstall with:

sudo apt remove kamite-bin

Generic

Download the latest release package from the Releases page and extract it to the location where you want to keep the program files. You can now launch Kamite with the bin/kamite executable inside the extracted directory.

An optional install.sh script is provided in the release package that can create a link to bin/kamite in the /usr/bin/ directory, as well as install a Desktop entry and program icons for application launchers. The files created by the installation script can be removed by running:

install.sh --uninstall

Windows

Download the latest release package from the Releases page and extract it to the location where you want to keep the program files. You can now launch Kamite using Kamite.exe inside the extracted directory.

Updating Kamite

When updating, please check the Changelog for breaking chages in the newer versions.

Linux Generic and Windows

To update, remove the old kamite directory and extract the new release package in its place.

Launching Kamite

Kamite can be launched:

  • Linux: either direclty using the bin/kamite executable in the program directory or through a desktop launcher (assuming the .desktop file was installed);

  • Windows: using one of: Kamite.exe, Kamite.com (provides console output), or Kamite (Debug).bat (enables Debug mode and provides console output).

Besides the config file, Kamite supports configuration through launch options. See Launch options.

Kamite’s main user interface is a webpage served by a local server that should be opened in your web browser. The default address is http://localhost:4110.

The web client is only guaranteed to work without issues on the most recent versions of Firefox and Chrome.

Upon launch, Kamite will, by default: 1) navigate to the above address in the default web browser, 2) open an auxiliary “Control” window, which lets you monitor the program’s status as well as exit it. Both those behaviours can be disabled by setting the config keys launchBrowser and controlWindow, respectively, to no. Disabling the latter is useful when Kamite is run from the command line, making the control window redundant.

Multiple config profiles can be prepared for Kamite and chosen between at the time of its launch. See Config profiles.

Troubleshooting

To better diagnose problems, launch Kamite from a console window with the launch option --debug. (On Windows, this can be done simply by double-clicking the provided Kamite (Debug).bat script). This will make Kamite print detailed error messages and add a Debug tab to the client (the bug icon) where image snapshots of OCR operations appear for inspection.

If something does not seem to work properly or is confusing in a way that is not dispelled by what is written in this README, do not hesitate to create a GitHub Issue.

User interface overview

Below are some non-obvious tips regarding the interface that might come useful.

  • Double-clicking on the current chunk activates the chunk edit mode. To exit it, either click away or press Ctrl + Enter.

  • The notebook can be temporarily resized by dragging its edge with the mouse (in order for the setting to persist, however, it must be set in the config file).

  • You can have incoming text chunk inserted into the current chunk instead of entirely replacing it. To do that, enter the chunk edit mode as described above and either: 1) place the caret where you want the incoming chunk to be inserted or 2) select the part of the current chunk that you wish replaced with it.

  • The character counter and the session timer can be frozen/paused and restarted by, respectively, short-clicking and long-clickng on them.

  • The session timer can be set up to pause automatically after a period of inactivity. For example, add sessionTimer.autoPause.after: 45s to auto-pause after 45 seconds of inactivity.

  • The session timer can be set to be paused initially after launching the program (sessionTimer.startPaused: yes to enable).

Text extraction

The first task when using Kamite is to get text from the immersion material into the program. In Kamite’s terminology, a single incoming piece of text is called a chunk.

The method of getting text chunks into Kamite will depend on the kind of the source material:

Anime/video text extraction

Extracting text chunks from video is by default supported through integration with the mpv video player. Primary video subtitles from mpv are treated by Kamite as incoming chunks. If secondary subtitles are present, they are treated as chunk translations.

To enable the connection between Kamite and mpv, the latter must be launched with the following exact value for the input-ipc-server parameter:

Linux:

mpv file.mkv --input-ipc-server=/tmp/kamite-mpvsocket

Windows (PowerShell):

C:\Program` Files\mpv\mpv file.mkv --input-ipc-server=\\.\pipe\kamite-mpvsocket

Alternatively, the line

input-ipc-server=<above_path>

can be put into the mpv config file.

In the former case, mpv will only be open for communication with Kamite when launched with the specified parameter. In the latter—it will be open always.

For more on the communication mechanism used, see the mpv reference for JSON IPC.

To run mpv with an external subtitle file, use the --sub-file launch parameter (it can be repeated for multiple subtitle files). To assign a given subtitle track as primary (assumed by Kamite to be the Japanese subtitles) and secondary (assumed to be the translations), respectively, use the --sid and --secondary-sid mpv launch parameters. Which subtitle IDs to specify can be glanced by pressing F9 in mpv while the video file is opened and the subtitles loaded.

[!NOTE] Subtitles hidden within mpv will still be recognized by Kamite.

[!WARNING] The subtitle extraction functionality will not work with subtitles that are stored as images, not as text.

See also: mpv reference: Subtitle options.

Below are excerpts from example scripts used to quickly launch an anime episode in mpv in such a way that it is immediately set up to work with Kamite.

Linux:

mpv "/path/to/video/"*"Some Anime Name"*"E$1"*".mkv" \ # Episode no. passed as an argument to the script
  --input-ipc-server=/tmp/kamite-mpvsocket \
  --sub-file="/path/to/external/subtitles/$1.jp.srt" \
  --sid=2 \ # ID of the Japanese subtitles provided externally
  --secondary-sid=1 \ # ID of the English subtitles embedded in the video file (to be used as translations)
  --secondary-sub-visibility=no \
  --save-position-on-quit \
  --profile=jpsub # An optional custom profile that can set a special subtitle font and size, etc. It must be defined separately in the mpv config file: see https://mpv.io/manual/stable/#profiles

Windows (PowerShell):

# See the Linux example above for more information
param ([String] $ep) # $ep will be replaced with the first parameter to the
                     # script (here assumed to be the episode number)
C:\Program` Files\mpv\mpv "\path\to\video\*Some Anime Name*E$ep.mkv" `
  --input-ipc-server=\\.\pipe\kamite-mpvsocket `
  --sub-file="\path\to\external\subtitles\$ep.jp.srt" `
  --sid=2 `
  --secondary-sid=1 `
  --secondary-sub-visibility=no `
  --save-position-on-quit `
  --profile=jpsub

Tip: Auto-pausing at the start/end of subtitles

This can be useful for beginners whose comprehension is not yet at the level needed to follow even simple video content at natural pace. The sub-pause-advanced mpv script can be used for this purpose. Here are some example learning workflows enabled by that script:

  • Pause at the end of each Japanese subtitle in order to read it in Kamite. (Useful when you cannot read proper texts yet and it is advantageous to substitute with short anime dialogues.)

  • Do the above, but only when you stumble while reading the subtitle live and you explicitly request pause at the end of the subtitle by pressing a defined request key. (Useful once you progress to the level at which you can often read easier lines live.)

  • Pause at the start of each {known language} subtitle and automatically unpause after the time sufficient to read it (calculated by the script based on the subtitle’s length), so that you can then fully concentrate on understandning the Japanese spoken line, with the comprehension “training wheel” in the form of knowing roughly what it’s supposed to say. (Useful for beginner listening practice.)

For more information, see the script’s README in its GitHub repository linked above.

Kamite can be useful even when viewing media without Japanese subtitles, for example as an area where heard words and phrases can be typed in and looked up.

When viewing media with translated subtitles only, Kamite can be instructed to treat them as translations for unknown chunks and display them as such, by enabling “Translation-only mode” in the Settings tab or by launching with the config key chunk.translationOnlyMode set to yes.

Related Wiki sections:

Manga/visual text extraction

Kamite integrates with several alternative OCR (Optical Character Recognition) solutions to enable visual extraction of text from manga pages and other content displayed on screen. The available OCR engines are:

“Manga OCR” in either variant is the recommended choice for manga. The online version requires practically no special setup, but involves screenshots of portions of your screen being sent to a third party. The local version, on the other hand, requires a somewhat involved setup and extra system resources.

By default, OCR is disabled. The necessary setup steps are:

  1. Set the config key ocr.engine to one of: mangaocr_online, mangaocr, hiveocr_online, easyocr_online, ocrspace, or tesseract.

  2. Set up the selected engine:

  3. (Linux/Xorg and wlroots platforms only) Set up extra OCR dependencies

Setting up “Manga OCR” Online

[!WARNING] The “Manga OCR” Online engine depends on a third-party online service (a Hugging Face Space by Detomo)—using it involves sending screenshots of portions of your screen to third parties. (See Privacy)

The “Manga OCR” Online does not require any extra setup.

Remember to set up extra OCR dependencies if necessary, and to launch Kamite with the config key ocr.engine set to mangaocr_online.

Setting up “Manga OCR” (Local)

[!NOTE] “Manga OCR” will use up to 2.5 GB of disk space. During launch, it will use up to 1 GB of additional RAM.

Recommended option: installation using pipx

  1. Install Python and pip (the Windows Python installer already includes pip)

  2. Install pipx

    Windows:

    C:\Users\<user>\AppData\Local\Programs\Python\Python310\python.exe -m pip
install --user pipx

    The python.exe location will vary depending on the installer and the installation options.

  3. Install manga-ocr

    Linux:

    pipx install manga-ocr

    Windows:

    C:\Users\<user>\AppData\Local\Programs\Python\Python310\python.exe -m pipx
install manga-ocr

Kamite will now be able to use “Manga OCR”. On the first launch of Kamite with ocr.engine set to mangaocr, “Manga OCR” will take some time to download its model (around 450 MB). If there are issues, try running the (Linux) manga_ocr, (Windows) C:\Users\<user>\.local\pipx\venvs\manga-ocr\Scripts\manga_ocr.exe executable installed by pipx and examining its output.

Deinstallation

  1. Run pipx uninstall

    Linux:

    pipx uninstall manga-ocr

    Windows:

    C:\Users\<user>\AppData\Local\Programs\Python\Python310\python.exe -m pipx
uninstall manga-ocr

  2. Delete leftover Hugging Face Hub files:

(Linux) ~/.cache/huggingface/hub/models--kha-white--manga-ocr-base or a ~450 MB file in ~/.cache/huggingface/transformers/.

(Windows) C:\Users\<user>\.cache\huggingface\hub\models--kha-white--manga-ocr-base or a ~450 MB file in C:\Users\<user>\.cache\huggingface\transformers.

Troubleshooting “pipx "Manga OCR" installation absent…”

If pipx did not install to the default path expected by Kamite, you will have to specify the path manually in the config file:

ocr: {
  mangaocr: {
    # Linux default
    pythonPath: "/home/<user>/.local/pipx/venvs/manga-ocr/bin/python"
    # Windows default
    pythonPath: """C:\Users\<user>\.local\pipx\venvs\manga-ocr\Scripts\python.exe"""
  }
}

The above paths are the defaults, which you will need to modify according to the output you get from running

Linux:

pipx list

Windows:

C:\Users\<user>\AppData\Local\Programs\Python\Python310\python.exe -m pipx
list
Custom installation

If you install “Manga OCR” not through pipx, you will need to manually specify a path to a Python main executable (or a wrapper for it) that runs within an environment where the manga_ocr module is available. For example, if installed globally and the system Python executable is on PATH under the name python, then the appropriate configuration will be simply:

ocr: {
  mangaocr: {
    pythonPath: python
  }
}

[!NOTE] After deinstallation, there will be a ~450 MB leftover model file in (Linux) ~/.cache/huggingface/transformers/, (Windows) C:\Users\<user>\.cache\huggingface\transformers.

Setting up Hive OCR Online

[!WARNING] The Hive OCR Online engine depends on a third-party online service (a Hugging Face Space by seaoctopusredchicken)—using it involves sending screenshots of portions of your screen to third parties. (See Privacy)

The Hive OCR Online engine does not require any extra setup.

Remember to set up extra OCR dependencies if necessary, and to launch Kamite with the config key ocr.engine set to hiveocr_online.

Setting up EasyOCR Online

[!WARNING] The EasyOCR Online engine depends on a third-party online service (a Hugging Face Space by tomofi)—using it involves sending screenshots of portions of your screen to third parties. (See Privacy)

The EasyOCR Online engine does not require any extra setup.

Remember to set up extra OCR dependencies if necessary, and to launch Kamite with the config key ocr.engine set to easyocr_online.

Setting up OCR.space

[!WARNING] OCR.space is an online service—using it involves sending screenshots of portions of your screen to a third party. (See Privacy)

The usage of the OCR.space free API is limited. The limits are defined by the provider as “Requests/month: 25000, Rate Limit: 500 calls/DAY”.

  1. Register for a free API key

    Fill the form at https://ocr.space/ocrapi/freekey. You will need to provide your email address, to which the key will be sent.

  2. Put the API key in Kamite’s config file:

    [!WARNING] This is unsafe plain-text storage. Do not do this if you deem your key too sensitive for this kind of storage.

    secrets: {
  ocrspace: "THE KEY GOES HERE"
}

  3. (Optional—for games) Change the subengine from the default (“1”)

    The OCR.space service itself provides multiple OCR engines. Japanese is supported by engines “1” and “3”. Kamite uses engine “1” by default. If you wish to use engine “3”, specify the following configuration:

    ocr: {
  ocrspace: {
    engine: 3
  }
}

    Engine “3” is useless for manga and slower than engine “1”, but it may be more accurate for, e.g., particular games.

Remember to set up extra OCR dependencies if necessary, and to launch Kamite with the config key ocr.engine set to ocrspace.

Setting up Tesseract OCR

  1. Install Tesseract

    Linux: Tesseract is available in the default repositories of most distributions. For example, under tesseract-ocr on Ubuntu or under tesseract on Arch Linux.

    Windows: It is recommended to use the installer provided by UB Mannheim.

  2. Install Tesseract models selected for use with Kamite

    Download tesseract_traineddata_jpn_Kamite.zip and extract the .traineddata files from the archive directly into Tesseract’s tessdata directory:

    • Linux: usually /usr/[local/]share/tessdata or /usr/share/tesseract-ocr/<VERSION>/tessdata;

    • Windows: the default for the UB Mannheim installer is C:\Program Files\Tesseract-OCR\tessdata.

  3. If Tesseract is not available on PATH under the executable name tesseract (which it will not on Windows), set the config key ocr.tesseract.path to its executable’s path:

    Windows (UB Mannheim installer default):

    ocr: {
  tesseract: {
    path: """C:\Program Files\Tesseract-OCR\tesseract.exe""" # Note the triple quotes
  }
}

Once the setup has been completed, you can launch Kamite with the config key ocr.engine set to tesseract.

Setting up extra OCR dependencies

On some platforms, external programs are needed for text recognition related tasks. You need to install them on your own.

Linux/Xorg extra OCR dependencies
slop
Used for selecting a screen region or point.
Linux/wlroots extra OCR dependencies
slurp
Used for selecting a screen region or point.
grim
Used for taking screenshots for OCR.
wlrctl
(Optional) Necessary to trigger a mouse click for OCR Auto Block Instant mode.

OCR usage

Text recognition can be initiated by:

[!WARNING] On Linux / GNOME Wayland and Plasma Wayland, only the following of the above are available:

The OCR commands directly available to the user are the following:

Manual block OCR

OCR manual block button OCR manual block vertical button OCR manual block horizontal button

Select an area around a block of text and Kamite will OCR the area as is.

For the tesseract engine, this command has separate vertical and horizontal variants that must be chosen manually depending on the orientation of the text block.

Auto block OCR

OCR auto block button

Select a point within a block of text; Kamite will try to infer the extent of the block and then OCR the resulting area.

This should be good enough for > 90% of typical manga text blocks, but the block detection algorithm has a lot of room for improvement.

[!NOTE] (Linux/Xorg) On Xorg, the point selection mechanism cannot be restricted to just a point, meaning that when the mouse is pressed and dragged, a rectangle area will be selected instead of a point. If this happens, Kamite will consider the center of this area as the selected point.

Manual rotated block OCR

OCR manual block rotated button

Delimit a rotated block of text; Kamite will derotate the resulting area selection and OCR it.

The delimitation of a rotated block is made with three mouse clicks in determined spots, as shown in the following illustrations:

OCR rotated block reference

Clicks 1 and 2 must be made at the start and end of the initial edge of the text respectively. Click 3 can be anywhere along the closing edge (pictured as green above).

[!NOTE] The current implementation of rotated block OCR guesses the text orientation based on the rotation angle. This means the feature will fail in unusual cases, such as a block of horizontal text positioned vertically. The current assumption is that those cases are very rare, but if you find use-cases where they are not, please create a GitHub Issue so that the assumption can be updated and the implementation reconsidered.

Region OCR

OCR region button

Define a screen area in the config file and Kamite will OCR it as is using the configured OCR engine.

This is intended as an alternative for games that do not work with Textractor or Agent.

Here it is recommended to try OCR engines in the following order: Hive OCR, EasyOCR, OCR.space, and “Manga OCR” (in terms of the likelihood that they are going to be up to the task). (For OCR.space, be sure to try OCR.space engine “3” specifically).

Below is an illustration of setting up a region in the config file.

ocr: {
  regions: [
    ${REGIONS.exampleGameTextbox}
    ${REGIONS.someOtherRegionOmittedBelow}
  ]
}

REGIONS: [
  exampleGameTextbox: {
    # (1 character) This symbol will appear on the region’s button
    symbol: E

    # This description will appear in the region’s button tooltip
    description: "OCR Example Game's textbox"

    # The screen coordinates of the region’s top-left corner
    x: 350
    y: 800

    # The screen dimensions of the region
    width: 1000
    height: 150

    # Try to automatically narrow the region’s screenshot to just text before
    # OCR-ing it.
    # NOTE: The implementation of this function is currently very basic. It
    #       might not prove helpful in most use-cases
    autoNarrow: no
  }
]

keybindings: {
  global: {
    ocr: {
      region: [
        # Global keybinding for the above-defined region
        { symbol: E, key: meta shift E }
      ]
    }
  }
}
Obtaining region parameters

To obtain the desired region coordinates and dimensions, use Kamite’s Region Helper mode: Launch Kamite from console with the launch option --regionHelper, select the desired areas, and copy the resulting region specifications from the console output.

[!NOTE] (Windows) To get console output on Windows, you must launch Kamite using the Kamite.com executable, not Kamite.exe.

Region OCR quality

Some of the engines that can potentially handle this specific task work much more reliably when the screenshot they are provided with is narrowed to just the text. And since the current auto-narrowing algorithm is poor, for now it might be best—when necessary—to create separate regions for each possible line count of the target text box (i.e., first region encompassing just one line of text, second region encompassing the first and the second line, and so on) and choose between them on the fly.

See also: Config, Visual novel / game text extraction, Alternative software for visual novels / games.

Using alternative OCR variants (Tesseract OCR only)

When using the Tesseract OCR engine, Kamite executes Tesseract multiple times for each OCR request on modified versions of the input image. This is in order to somewhat compensate for the inferior results Tesseract gives.

When all the resulting text variants are identical, they are simply merged and presented as one. When there are differing alternative variants, however, the one determined the most likely to be the most accurate is displayed normally as the current chunk, whereas the remaining variants appear under a new OCR Variants tab placed at the left edge of the notebook’s tab bar.

Through the OCR Variants view, you can replace the current chunk with one of the other variants by clicking the Pick button next to the corresponding variant. You can also make text selections within both the current chunk and the variants, which, upon clicking the Pick button, will lead to a partial text replacement.

The OCR Variants view highlights characters that are unique to its variant to make it easier to identify potential replacements for a misrecognized character in the current chunk.

OCR directory watcher

Kamite can watch a specified directory for new/modified images and perform text recognition on them automatically. This is especially useful for platforms that do not support global OCR commands (Linux / GNOME Wayland and Plasma Wayland).

To enable the directory watcher, specify the directory path in the config file:

ocr: {
  watchDir: "/full/path/to/the/directory" # Linux variant
  watchDir: """C:\path\to\the\directory""" # Windows variant
}

This can be used in conjunction with a platform-specific screenshot tool, for example GNOME Screenshot on GNOME Wayland and Spectacle on Plasma Wayland.

Note that Kamite treats the entire input image as a single text block. Consequently, the provided images should be area screenshots containing just the text block to be recognized—not the entire screen or application window.

Web browser userscript for convenient OCR

Kamite One-Click OCR is a simple browser userscript that enables convenient shortcuts for issuing OCR commands to Kamite (useful when reading manga directly in the browser):

  • Middle-click on a text block to automatically recognize it.

  • Middle-hold-click anywhere within the webpage to initiate a manual block recognition selection.

To install the script, first install a userscript browser extension such as Violentmonkey or Tampermonkey, and then follow this link: Kamite One-Click OCR.user.js.

Then you need to specify on which websites the script will be active:

  1. Open your userscript browser extension.

  2. Click on the cog icon (“Open dashboard”).

  3. Find Kamite One-Click OCR and edit it:

    Tampermonkey: Click on the script name.
    Violentmonkey: Click the </> (“Edit”) icon below the script name.

  4. Switch to the Settings tab.

  5. Add user match rules for desired websites.

    Example configuration enabling the script on two specific websites:

    Enter under User matches (Tampermonkey) or @match rules (Violentmonkey):

    https://twitter.com/*
https://metaurl.app/*

On first use, you might be prompted to confirm that you allow the script to make web requests to Kamite’s backend server.

You can switch to using the right mouse button instead of the middle by going to the Code tab and changing const MOUSE_BUTTON = 1 to say = 2. Note that this will disable the browser context menu that normally opens on right-click.

Linux manga viewer with Kamite integration

Gomics-v is a manga viewer for Linux that includes simple Kamite integration:

  • Right-click on a text block to automatically recognize it.

  • Right-hold-click anywhere in the area where images are displayed to initiate a manual block recognition selection.

The integration must be enabled in Gomics-v under Preferences › Kamite.

Related Wiki sections:

Visual novel / game text extraction

Kamite can receive text extracted from other programs through Textractor and Agent text extractor programs:

There is also a RPG Maker MV/MZ plugin that can directly extract text from games using those engines: see contrib/KamiteSendRPGMaker.js (usage instructions are in the plugin file).

Other integrations can be created using the Command API.

[!TIP] For games that cannot be integrated using the above means, the Region OCR feature might prove an alternative.

Textractor integration

An extension for Textractor is available that sends lines of text extracted by it from other programs to Kamite.

To install the Kamite Send Textractor extension:

  1. Copy Kamite Send.xdll from either the extra/textractor/x86 (32-bit Textractor) or extra/textractor/x64 (64-bit Textractor) within the release package to the main Textractor directory (the one containing Textractor.exe).

    [!NOTE] If you installed Kamite from AUR, the files will be in /opt/kamite.

  2. In Textractor, press the Extensions button, right-click in the extension list, choose Add extension and select the Kamite Send.xdll file copied earlier.

For Textractor’s text processing extensions (such as Remove Repeated Characters) to be executed before sending the text to Kamite, position Kamite Send below them on the extensions list.

Want to prevent Textractor from inserting unnecessary hooks? Check out Textractor-HookAllowlist.

Changing the default Textractor extension endpoint

By default, the extension expects Kamite to listen for commands at the address localhost:4110. If it is running on a different address (e.g., because you have Textractor in a virtual machine and Kamite on the host), the extension needs to be configured correspondingly. To do so, create a Kamite Send.txt file next to Kamite Send.xdll in the Textractor directory, containing Kamite’s network address from the perspective of the machine running Textractor, e.g.:

host=192.0.0.10:4110

Agent integration

To receive text (and, optionally, machine translations) from the Agent text extractor:

  1. Add integrations.agent.enable: yes to config.

  2. In Agent, go to Translate tab and enable the option WebSocketServer (you can then browse to http://127.0.0.1:9001/ to confirm that it is enabled).

With this setup, Kamite should receive text from Agent.

[!TIP] (Linux) To hook Wine games, you can run the Windows version of Agent through Wine (you might need an older version of Wine, such as 6.14).

Related Wiki sections:

Clipboard

Text can be pasted from clipboard by pressing Ctrl + V in Kamite’s browser tab.

The Kamite browser client can automatically pick up clipboard text with a clipboard inserter browser extension (Firefox, Chrome) (assumes default extension settings).

Watcher script

Clipboard can also be watched automatically without the Clipboard Inserter extension, but with a clipboard watcher script that sends changed clipboard content to Kamite:

Custom source / alternative method text extraction

Custom sources and alternative text extraction methods can be integrated using the Command API.

Filtering and transforming incoming text

Kamite has the ability to discard incoming chunks that match a user-provided pattern and to perform text replacements on incoming chunks according to user-provided rules.

Both features are based on regular expressions, which can be learned using the following resources:

  • RegexOne — interactive regular expression tutorial.
  • regex101 — regular expression tester and explainer.

[!NOTE] The particular regular expression engine used by Kamite is Java’s, which has some particularities. Therefore, Java’s own regular expression reference might also come useful (note that regex101 can be instructed to use the Java flavour of regex through the menu on the left side of the page).

Filtering chunks

To enable the chunk filter, specify chunk reject patterns in the config (chunk.filter.rejectPatterns). The patterns are regular expressions against which incoming chunks will be tested. If any part of the chunk matches against any of the patterns, the entire chunk will be discarded. For exmple, the following configuration:

chunk: {
  filter.rejectPatterns: [
    "^Textractor"
    "(?s).{91}"
  ]
}

will discard both chunks beginning with the string of characters Textractor and chunks that are above 90 characters in length. The first of those filters is already included in the default config that is created automatically when no config file exists.

[!WARNING] Regular expressions containing the backslash (\) character must (generally) be specified within triple instead of single quotes (that is, """ instead of "). See the section Transforming chunks (following) for an example.

[!NOTE] Kamite will automatically reload the patterns once a config file is modified. There is no need to restart the program. The simplest way to test the filter is by pasting (Ctrl + V) prepared chunks into the client’s browser tab and observing whether they are allowed through.

Transforming chunks

To enable the chunk transformer, specify chunk transform rules in the config (chunk.transforms). The rules are objects consisting of: 1) a regular expression defining the replacement target, 2) a string specifying the replacement text. For example, the following configuration:

chunk: {
  transforms: [
    { replace:  """\R""", with: "" }
    { replace: ".+?「(.+?)」$", with: "$1" }
  ]
}

will remove (replace with an empty string) any line break characters as well as remove the names of speakers and quotation marks from dialogue lines formatted as <name>「<dialogue>」 (this will be replaced with just <dialogue>. The $1 in the with field is a special reference that will be replaced with the text captured in the first and only match group defined within the replace field). The rules will be applied in the specified order.

[!WARNING] Regular expressions containing the backslash (\) character must (generally) be specified within triple instead of single quotes (that is, """ instead of "). See the first transform in the example above for illustration.

[!NOTE] Kamite will automatically reload the transform definitions once a config file is modified. There is no need to restart the program. The simplest way to test the transforms is by pasting (Ctrl + V) prepared chunks into the client’s browser tab and observing whether they come out modified as expected.

Text use

What can be done with the text once it has been collected into the program?

Editing and transforming the text

Depending on the source and the extraction method, the text coming into the program might contain mistakes or other features that need to be changed for it to be usable further. Kamite offers various text editing and transformation features that might be useful for that task.

The main chunk text can be selected character by character using the mouse and transformed using actions that appear in the action palette near it, which are made available depending on the existence and the content of the selection. Characters can also be deleted by pressing Delete or Backspace.

To enter chunk edit mode, which allows for direct insertion of typed text, double-click the main area or press Ctrl + Enter (+ Shift to clear). To exit it, click away from the input field or press Ctrl + Enter.

Pop-up dictionary

The basic task around which Kamite is designed is using a third-party pop-up dictionary browser extension to learn through deciphering text you understand only partially. The examples of such extensions are Yomichan, rikaikun, and 10ten Japanese Reader.

For Yomichan, its website contains a basic installation and usage guide. Additional dictionaries and setup tips for can be found in other places, such as Animecards: Setting up Yomichan, TheMoeWay: Yomichan Setup Tutorial, or Anacreon DJT: Yomichan Frequency Dictionaries.

Lookups

You can use the text chunks within Kamite as inputs for lookups on external websites. The text that will be used as the lookup input normally corresponds to the selection in the chunk history window, which, absent user intervention, is the entire current chunk. If instead there is an active text selection within the current chunk, it is only the selection that will be used for lookups.

By default, Kamite embeds the following sites for quick lookups: https://www.deepl.com/, https://ichi.moe/, https://jpdb.io/. It also by default provides a button for Google Images lookup in a new browser tab.

[!TIP] A faster and more robust alternative to using DeepL as a Kamite lookup is to use a scripting-friendly DeepL API client in combination with Kamite’s custom command mechanism.
The Wiki contains a Linux recipe for how to use the DeepLX program to get DeepL translations into Kamite’s chunk translation component on button click.

Optional userscripts are provided for the embedded sites that modify their look when embedded into Kamite to improve their user experience. To install or update the scripts, first install a userscript browser extension such as Violentmonkey or Tampermonkey, and then open the links below and follow the prompt.

DeepL and jpdb require yet another intervention to work. A browser extension to slightly modify the responses from those sites is required to make them available for embedding.

  1. Install the SimpleModifyHeaders browser extension (Firefox, Chrome)

  2. In the extension’s configuration screen, set the field Url Patterns to https://www.deepl.com/*;https://jpdb.io/*

  3. Below that, enter the following parameters (for DeepL):

    Field Value
    Action Delete
    Header Field Name content-security-policy
    Apply on Response

  4. Click New line and enter the following into the second line of fields (for jpdb):

    Field Value
    Action Delete
    Header Field Name X-Frame-Options
    Apply on Response

  5. Click Save.

  6. Click START in the top-right corner.

    DeepL and jpdb embeds should now work.

Alternatively, you can have DeepL and jpdb lookups open in a new browser tab by simply adding the line newTab: yes to the corresponding lookup config entries.

Custom lookups

Kamite supports adding custom lookups. Below is an illustration of the process using the website Immersion Kit as an example lookup target.

  1. Add an object describing the target to the LOOKUP_TARGETS object in the main config file:

    LOOKUP_TARGETS: {
  # This key can be anything
  immersionKit: {
    # (1-3 characters) The symbol will appear on the lookup button
    symbol: IMK

    # The name will appear in the lookup button’s tooltip
    name: Immersion Kit

    # {} is a placeholder that will be replaced with the lookup text
    url: "https://www.immersionkit.com/dictionary?keyword={}"

    # (Optional) Whether to open the lookup in a new browser tab or embedded
    newTab: yes
  }
}

  2. Add the above-created lookup target to the lookup.targets list:

    lookup: {
  targets: [
    ${LOOKUP_TARGETS.immersionKit} # Has to match the key defined above
  ]
}

Auto-generated furigana

Kamite can add auto-generated furigana annotations to the current chunk. Please keep in mind that the auto-generated annotations will frequently be incorrect. To enable this feature:

  1. Get the necessary library file (a morphological analyzer with an included dictionary).

    Download kuromoji-unidic-kanaaccent-e18ff911fd.jar and put it directly in:

    • Linux: either $XDG_DATA_HOME/kamite or, if that is not set, $HOME/.local/share/kamite;

    • Windows: same directory as the config file (usually C:\Users\<user>\AppData\Roaming\kamite).

  2. Set chunk.furigana.enable: yes in the config file and start Kamite; or, to enable temporarily, start Kamite and switch on the “Enable furigana” setting in the Settings tab.

There is an option to hide the generated furigana annotations until the base text is hovered with the mouse (chunk.furigana.conceal: yes).

Saving text to a file for external use

Kamite can be instructed to save all chunks that appear in the client into a text file for further use (for example, for purposes of statistical analysis of the texts you read). To enable chunk logging, specify the path (in the config) of the directory where the text files ought to be saved:

chunk: {
  log.dir: "/path/to/the/directory" # Linux variant
  log.dir: """C:\path\to\the\directory""" # Windows variant
}

Kamite will then create a dated log text file in that directory for each session (that is, one each time the program is started), to which chunks appearing in the client will be appended live, one chunk per line (note: to that end, line breaks within chunks will be replaced with their literal representation, \n, so that actual line break characters demarcate chunks).

To disable logging, either remove the log.dir definition before starting Kamite or simply comment it out by putting the # character in front of it.

Custom text use

Text from Kamite can be forwarded to external programs using Custom commands (see below). Example uses:

Custom commands (Launching external executables)

You can add buttons to the command palette that execute specified system executables. Those executables can optionally receive data from Kamite as arguments.

Below is an excerpt from a config file illustrating how to define a custom command.

commands: {
  custom: [
    ${CUSTOM_COMMANDS.exampleCustomCommand}
    ${CUSTOM_COMMANDS.anotherExampleCustomCommand}
  ]
}

CUSTOM_COMMANDS: {
  exampleCustomCommand: {
    # (1-3 characters) The symbol that will appear on the command's button
    symbol: CMD

    # The name that will appear in the command button's tooltip
    name: Example command

    # (String list) The system command to execute.
    # The first element in the list is the executable, the rest are the
    # arguments. The arguments can be specific placeholders to be filled by
    # Kamite at execution time.
    command: ["/path/to/the/executable.sh", "some argument", "{effectiveText}"]
  }
}

[!NOTE] (Windows) To execute a PowerShell script, set command to, e.g., ["powershell.exe", """C:\path\to\the\script.ps1""", "first argument"] (note the triple quotes). The execution of PowerShell scripts is disabled by default in the system. To enable it, start PowerShell with the Run as Administrator option and execute the command Set-ExecutionPolicy RemoteSigned. Be aware that this lowers system security.

The supported placeholder arguments for custom commands are:

{effectiveText}
Equivalent to the text used for lookups at the given moment. If there is a selection in the current chunk, just the selection will be used. Otherwise, the text selected in the chunk history view will be used.

{originalEffectiveText}
Same as {effectiveText}, except yields the original text for chunks that have it. Chunks have original text if their text was modified by Kamite’s automatic correction procedure or user-provided transforms.

Keyboard shortcuts

Kamite is designed under the assumption that the user’s main activity in the program will be hovering over the current chunk text with the mouse for dictionary lookups. Support for keyboard shortcuts is therefore very basic, for the time being.

Client-only keyboard shortcuts

The following shortcuts apply only within the Kamite’s browser tab.

Backspace, Delete
Delete selected text from the current chunk.

Ctrl + A
Select the entire current chunk text.

Ctrl + C
Copy text depending on what is selected in the current chunk, the current chunk translation and chunk history.

Ctrl + Alt + C
Copy original text depending on what is selected in the current chunk or chunk history. Original text is available in the case when Kamite’s built-in corrections or user-specified transforms have been applied to the text. Otherwise, this acts the same as Ctrl+c.

Ctrl + V
Paste text from clipboard.

Ctrl + Enter
Enter chunk edit mode.

Ctrl + Shift + Enter
Enter chunk edit mode erasing the current chunk text.

Chunk edit mode keyboard shortcuts

Ctrl+Enter
Exit chunk edit mode.

Media player keyboard shortcuts (client)

The following shortcuts work only when a media player is connected.

Space
Pause/unpause the playback.

◄
Seek -1 seconds.

Alt + ◄
Seek to the start of the current subtitle.

►
Seek +1 seconds.

Global keyboard shortcuts

Linux/Xorg and Windows

Below is an excerpt from a config file illustrating how to set up global keyboard shortcuts and what actions are available for binding.

keybindings: {
  global: {
    ocr: {
      manualBlock: …
      manualBlockRotated: …
      autoBlock: … # Instant detection under mouse cursor
      autoBlockSelect: … # Must click to select a point

      region: [
        { symbol: …, key: … }
      ]
    }
  }
}

The values must be strings in the Java Swing’s getKeyStroke() format describing the key combinations to be assigned to the corresponding actions. Some examples of such strings are:

  • meta S (means Win + S),
  • ctrl shift 1,
  • INSERT.

Linux/wlroots

Global shortcuts on the Linux/wlroots platform must be created manually with the use of the Command API. The details of the process will differ depending on the compositor used.

Below is an example of setting up global keyboard shortcuts in Sway’s config file:

bindsym $mod+s exec "dbus-send --type=method_call --dest=io.github.kamitejp /Receiver io.github.kamitejp.Receiver.command string:'ocr_manual-block'"
bindsym $mod+d exec "dbus-send --type=method_call --dest=io.github.kamitejp /Receiver io.github.kamitejp.Receiver.command string:'ocr_auto-block'"

Launch options

Windows: To launch Kamite with extra options, either: 1) create a shortcut to Kamite.exe, open its Properties, and append the options to the Target string after …\Kamite.exe (e.g., …\Kamite.exe[space]--profile=myprofile) or 2) append them when invoking Kamite.exe or Kamite.com from a console window.

First, there are the basic launch options that are read before the config file:

--help
Prints the usage message and exits.

--version
Prints the program version and exits.

--profile=<profile-id>[,<second-profile-id>]+
Instructs Kamite to load the profile config files with the specified IDs. See Config profiles.

--debug[=all]
Instructs Kamite to print debug messages to console as well as to enable certain extra features in the client that are useful for debugging purporses.
Setting the value to all will instruct Kamite to additionally display debug messages from the third-party dependencies.

--regionHelper
Launches Kamite in Region Helper mode used to obtain coordinates of screen regions. Useful for Region OCR.

Beyond that, all simple config settings can be overridden through correspondingly-named launch options. For example, launching Kamite with the option --ocr.engine=none will give that value precedence over whatever the value of the key ocr.engine in the effective config is.

Config

Kamite is configured through a config file placed in the directory:

  • Linux: either $XDG_CONFIG_HOME/kamite or, if that is not set, $HOME/.config/kamite;

  • Windows: %AppData%\kamite (usually C:\Users\<user>\AppData\Roaming\kamite).

The main config file must have the name config.hocon.

The config file’s format is HOCON, which is a superset of JSON.

A default config file is automatically placed by Kamite in the expected directory if absent on launch.

[!WARNING] (Windows) On unupdated Windows 10 installations as well as earlier Windows versions, the system Notepad will not be able to properly display the config file. In that case, please use another text editor.

[!WARNING] When providing values that containin backslashes (\), for example Windows paths, you must enclose them within triple quote marks ("""). For example: ocr.tesseract.path: """C:\Program Files\Tesseract-OCR\tesseract.exe""".

Live reload

The config will be automatically reloaded if one of the files is modified. However, only select changes will be applied immediately. To apply others, Kamite will still have to be restarted. The config keys for which the relevant changes are applied immediately are marked with [RELOADABLE] in the following section.

Full config example

Below is an example config file illustrating all the possible options with their default values (… represents no default value):

# Whether to launch Kamite with the control window or in console only
controlWindow: yes

# Whether to open the client in the default web browser upon launching Kamite
launchBrowser: yes

chunk: {
  # (Milliseconds) The minimum allowed delay between successive incoming chunks
  # before they begin to be throttled
  throttleMS: 1000

  # [RELAODABLE] Perform slight formatting corrections on incoming chunks
  correct: yes

  # [RELOADABLE] Allow flashing backgrounds of chunk texts in the client's
  # interface when they are changed or copied
  flash: yes

  # [RELOADABLE] Treat incoming chunks as translations and create a new chunk
  # for each translation. Useful when watching media with just the translation
  # subtitles
  translationOnlyMode: no

  # [RELOADABLE]
  furigana: {
    # Add auto-generated furigana annotations to the current chunk. Note that
    # this feature requires an extra download (see the Auto-generated furigana
    # section in the README).
    # WARNING: The auto-generated furigana will frequently be incorrect
    enable: no

    # Hide the furigana annotations behind rectangles, reveal on mouse hover
    conceal: no
  }

  # [RELOADABLE]
  log: {
    # A path of the directory where text files with chunks appearing in the
    # client should be saved. Chunk logging is disabled when this key is absent
    dir: …
  }

  # [RELOADABLE]
  filter: {
    # A comma-delimited list of strings representing regular expressions against
    # which incoming chunks will be matched. If there is at least one match, the
    # matching chunk will be rejected entirely. See the "Filtering chunks"
    # section of the Readme
    rejectPatterns: […]
  }

  # [RELOADABLE] A *list* of objects describing text replacement rules for
  # incoming chunks. See the "Transforming chunks" section of the Readme
  transforms: [
    {
      # A regular expression whose match will be replaced in incoming chunk texts.
      # Enclose in triple quotes to avoid problems with special characters
      replace: …

      # The replacement text. Supports match groups (`$1` stands in for the 1st
      # capture group in the `replace` regular expression, and so on)
      with: …
    }
  ]
}

commands: {
  # [RELOADABLE]
  player: {
    # Show extra media player controls (seek -+1 second, seek to the start of
    # the current subtitle)
    showExtra: yes
  }

  # [RELOADABLE] A *list* of custom commands that allow launching system
  # executables through buttons in Kamite's command palette. See the
  # "Custom commands" section of the Readme
  custom: [
    {
      # (1-3 characters) The symbol that will appear on the command's button
      symbol: …

      # The name that will appear in the command button's tooltip
      name: …

      # (String list) The system command to execute. Can contain certain
      # placeholders which will be filled in by Kamite at the moment of the
      # command’s activation
      command: …
    }
  ]
}

# For development only. See the "Development" page in the Wiki
dev: {
  serveStaticInDevMode: no
}

integrations: {
  # [RELOADABLE]
  agent: {
    # Whether to receive text from a running instance of Agent (the option
    # `WebSocketServer` must be enabled in Agent)
    enable: no

    # The address of the Agent's WebSocket server through which text is
    # received by Kamite. (There should be no need to override the default
    # value unless you run Agent and Kamite on different network hosts)
    host: "127.0.0.1:9001"
  }
}

keybindings: {
  # Global keybindings. See the "Keyboard shortcuts" section of the Readme.
  # WARNING: Not all platforms support global keybindings
  global: {
    ocr: {
      # A key combination to assign to the command. See the "Keyboard shortcuts"
      # section of the Readme for the format specification.
      manualBlock: …
      manualBlockRotated: …
      autoBlock: … # Instant detection under mouse cursor
      autoBlockSelect: … # Must click to select a point

      # Bindings for user-defined OCR regions by symbol
      region: [
        { symbol: …, key: … }
      ]
    }
  }
}

# [RELOADABLE]
lookup: {
  # A list of lookup targets that will be displayed in the notebook’s tab bar.
  # Consult the default config for further illustration of configuring lookups
  targets: [
    {
      symbol: … # (1-3 characters) The symbol will appear on the lookup button
      name: … # The name will appear in the lookup button's tooltip

      # The lookup's URL that should contain the placeholder {} which will be
      # replaced with the lookup text by Kamite
      url: …

      # Open the lookup in a new browser tab or embed it into Kamite’s notebook
      newTab: no
    }
  ]
}

ocr: {
  # The OCR engine to use: none, tesseract, mangaocr, mangaocr_online, ocrspace,
  #                        easyocr_online, hiveocr_online
  engine: none
  # (Directory path) Watch the specified directory for new/modified images and
  # OCR them automatically
  watchDir: …

  tesseract: {
    # (File path) The path to Tesseract’s executable
    path: "tesseract"
  }
  
  mangaocr: {
    # (File path) A path to a python executable that provides access to the
    # `manga_ocr` module. If absent, a system-dependent default is used which
    # assumes that manga-ocr was installed through pipx into the default
    # location
    pythonPath: …
  }

  ocrspace: {
    # (1, 3) The OCR.space engine to use (see # https://ocr.space/OCRAPI#ocrengine)
    engine: 1
  }

  # [RELOADABLE] A *list* of OCR regions, for each of which a region recognition
  # command button will be displayed in the command palette. See the
  # "OCR region" section of the Readme for details
  regions: [
    {
      symbol: … # (1 character) The symbol will appear on the region’s button
      description: … # The description will appear in the region’s button tooltip

      # (Numbers) The screen coordinates of the regions’s top-left corner
      x: …
      y: …

      # (Numbers) The screen dimensions of the region
      width: …
      height: …

      # Try to automatically narrow the region’s screenshot to just text before
      # OCR-ing it.
      # NOTE: The implementation of this function is currently very basic. It
      #       might not prove helpful in most use-cases
      autoNarrow: no
    }
  ]
}

server: {
  # The port on which to run the client and the API
  port: 4110
}

# [RELOADABLE]
sessionTimer: {
  autoPause: {
    # Whether to enable auto-pausing the session timer after a period of
    # inactivity. (Overriding this particular option is only useful for
    # disabling auto-pausing from a profile config or from the command-line in
    # case the `after` option has been set in an earlier config)
    enable: yes

    # (Time duration) The inactivity time interval after which to pause the
    # session timer. Example valid values: `30s`, `5m`. (If this option is set,
    # auto-pausing can still be disabled by setting `enable` to `no`)
    after: …
  }

  # Whether to pause the session timer initially at program launch
  startPaused: no
}

# [RELOADABLE]
ui: {
  # The client's user interface layout (standard, standard_flipped)
  layout: standard

  # Hide non-essential interface elements until mouse hover
  focusMode: no

  notebook: {
    # Automatically collapse the client's notebook to just its tab bar,
    # expanding only when it's being interacted with
    collapse: no

    # (25-90) The height of the client's notebook as a percentage of the total
    # browser inner window height
    height: 60
  }
}

update: {
  # Whether to check for updates on program launch and, if one is available,
  # print a notification to the Control GUI and the console
  check: yes
}

# Secrets used for authentication to third-party services.
# WARNING: This is unsafe plain-text storage. Do not put data here that you deem
#          too sensitive for this kind of storage
secrets: {
  # The OCR.space API key
  ocrspace: …
}

Substitution (variables)

The config format supports substitution. You can define your own keys and then use them when specifying values for the keys read by Kamite, listed above. For example, the following configuration:

chunk.log.dir: ${MY_HOME_DIRECTORY}"documents/kamite-chunk-logs"
ocr.mangaocr.pythonPath: ${MY_HOME_DIRECTORY}".local/pipx/venvs/manga-ocr/bin/python"

MY_HOME_DIRECTORY: "/home/my-username/"

will be resolved to:

chunk.log.dir: "/home/my-username/documents/kamite-chunk-logs"
ocr.mangaocr.pythonPath: "/home/my-username/.local/pipx/venvs/manga-ocr/bin/python"

Kamite will warn you if you specify an unknown key in a config file, since mistyping their names is a common source of mistakes. But it will at the same time ignore keys whose names start with a capital letter, under the assumption they are your own keys, like MY_HOME_DIRECTORY in the example above. For this reason, it is advised to name your own keys using capital letters to avoid confusion and false positive warnings.

Config profiles

Kamite supports having a set of different config files for different use-cases.

Creating in the config directory a file named config.example-profile.hocon and then launching Kamite with the launch option --profile=example-profile will instruct Kamite to take into account both the main config.hocon file and the config.example-profile.hocon file, with values from the latter taking precedence.

The values from all config files are resolved within a common context, which means that the files share the objects defined in them. For example, this works as expected:

config.hocon

lookup: {
  targets: [
    ${LOOKUP_TARGETS.deepl}
    ${LOOKUP_TARGETS.jpdb}
    ${LOOKUP_TARGETS.googleImages}
  ]
}

LOOKUP_TARGETS: {
  deepl: {
    symbol: DEP
    name: DeepL
    url: "https://www.deepl.com/translator#ja/en/{}"
  }
  jpdb: {
    symbol: JDB
    name: jpdb
    url: "https://jpdb.io/search?q={}"
  }
  googleImages: {
    symbol: GLI
    name: Google Images
    url: "https://www.google.com/search?q={}&tbm=isch"
    newTab: yes
  }
}

config.no-translations.hocon

lookup: {
  targets: [
    ${LOOKUP_TARGETS.googleImages}
  ]
}

Given the above configuration, launching normally shows deepl, jpdb, and googleImages lookups, whereas launching with --profile=no-translations shows only googleImages lookup.

Combined profile (multiple extra config files)

Several extra config files can be loaded at once. To achieve that, specify multiple profile IDs with the --profile launch option, separating them by commas, e.g., --profile=first-profile,second-profile. The values from the leftmost profile’s config file will have the highest precedence.

Style customization

If present, Kamite loads the custom.css file from the root of the config directory, which can contain style customizations for the client. Refer to your web browser’s Inspector view to see what CSS variables can be overridden and what ids and class names the various elements on the page can be referred to with.

Below is an example custom.css file to illustrate how to go about the customization.

:root {
  --color-bg: navy;
  --color-bg-hl: navy;
  --color-bg2: blue;
  --color-fg: fuchsia;
}
#character-counter * {
  font-size: 28px;
}
.notebook-tab {
  border: 3px dashed green;
}
#chunk-label:after {
  content: "Design is my passion edition";
  color: yellow;
}

The root client element is assigned CSS classes profile-X, where X stands for each profile specified at launch. Style customizations can therefore be defined per profile. For example, the following custom.css:

.profile-vn {
  --color-bg: purple;
}

will recolor the background to purple only when the program was launched with the profile named vn.

Styling recipes

Kamite black theme

Examples of ready-made style customizations, including a black color theme, can be found on the Styling recipes Wiki page.

Command API

Kamite can receive commands sent programmatically, which makes it simple to it integrate with all kinds of custom workflows.

Available commands are distinguished by command kind, which is made up of two segments: command group and command name. For example, kind ocr_region corresponds to the group ocr and the name region.

The command parameters are required unless they are marked as optional or a default value is specified.

Sending commands

Below are examples illustrating the available means of issuing commands to Kamite through its API.

DBus (Linux)

dbus-send --type=method_call \
  --dest=io.github.kamitejp /Receiver io.github.kamitejp.Receiver.command
  string:"chunk_show" \ # command kind, i.e. "{command_group}_{command_name}"
  string:'{"chunk": "This is the chunk text"}' # JSON object of command parameters

HTTP

Linux:

curl -i -X POST \
  -d '{"chunk": "This is the chunk text"}' \
  localhost:4110/cmd/chunk/show # {kamite_host}/cmd/{command_group}/{command_name}

Windows (PowerShell):

$body = @{
  "chunk" = "This is the chunk test"
}
Invoke-RestMethod -Method 'Post' `
  -Uri 'http://localhost:4110/cmd/chunk/show' `
  -Body "$($body | ConvertTo-Json)"

Command listing

See the Command API listing page in the Wiki.

Privacy

Kamite never saves your data to disk.

Kamite only sends data through the internet in the following circumstances:

  • When update.check is not set to no, a connection to github.com is made on every launch in order to check for the availability of a newer version of Kamite.

  • When ocr.engine is set to mangaocr_online, screenshots of portions of your screen are sent to a Hugging Face Space by detomo for text recognition.

  • When ocr.engine is set to hiveocr_online, screenshots of portions of your screen are sent to a Hugging Face Space by seaoctopusredchicken for text recognition.

  • When ocr.engine is set to easyocr_online, screenshots of portions of your screen are sent to a Hugging Face Space by tomofi for text recognition.

  • When ocr.engine is set to ocrspace, screenshots of portions of your screen are sent to OCR.space for text recognition.

Screenshots sent to third parties can be inspected in the client’s Debug tab (see Troubleshooting). They are labelled as “Remote OCR”.

Development

Kamite is built mainly with Java and SolidJS (TypeScript).

For more information see the Development page in the Wiki.

License

Kamite
Copyright (C) 2021-2023 Piotr Grabowski

This program is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

You should have received a copy of the GNU Affero General Public License along with this program. If not, see https://www.gnu.org/licenses/.

Third-party components

The Kamite repository and the Kamite release package include the following third-party components (as well as their subcomponents):

Component License
Apache Log4j Apache-2.0
hypfvieh/dbus-java MIT
Jackson Apache-2.0
Javalin Apache-2.0
jkeymaster LGPL-3.0
Java Native Access (JNA) Apache-2.0
jsoup MIT
Kuromoji UniDic Kana Accent Apache-2.0
Noto Sans Japanese OFL
Roboto Apache-2.0
SLF4J MIT
Solid MIT
Solid Styled Components MIT
tscfg Apache-2.0
Typesafe Config Apache-2.0

Kamite also includes components adapted from third-party code. See https://github.com/fauu/Kamite/search?q=adapted for the listing, including the original license notices.

Open Source Agenda is not affiliated with "Kamite" Project. README Source: fauu/Kamite
Stars
92
Open Issues
2
Last Commit
2 months ago
Repository
fauu/Kamite
License
AGPL-3.0
Tags
Anime Deepl Immersion Japanese Japanese Language Japanese Study Java Language Learning Languages Linux Manga Mpv Ocr Solidjs Tesseract Textractor Typescript Visual Novel Visual Novels Yomichan

Open Source Agenda Badge

Open Source Agenda Rating
Submit Review Review Your Favorite Project
Submit Resource Articles, Courses, Videos
Submit Article Submit a post to our blog

From the blog

Dec 11, 2022

How to Choose Which Programming Language to Learn First?

From the blog

Dec 11, 2022

How to Choose Which Programming Language to Learn First?