Image/Manga Translator

Translate texts in manga/images.
中文说明 | Change Log
Join us on discord https://discord.gg/Ak8APNy4vb

Some manga/images will never be translated, therefore this project is born.
Primarily designed for translating Japanese text, but also supports Chinese, English and Korean.
Supports inpainting and text rendering.
Successor to https://github.com/PatchyVideo/MMDOCR-HighPerformance

This is a hobby project, you are welcome to contribute!
Currently this only a simple demo, many imperfections exist, we need your support to make this project better!

Support Us

GPU server is not cheap, please consider to donate to us.

Ko-fi: https://ko-fi.com/voilelabs
Patreon: https://www.patreon.com/voilelabs
爱发电: https://afdian.net/@voilelabs

Online Demo

Official Demo (by zyddnys): https://touhou.ai/imgtrans/
Browser Userscript (by QiroNT): https://greasyfork.org/scripts/437569

Note this may not work sometimes due to stupid google gcp kept restarting my instance. In that case you can wait for me to restart the service, which may take up to 24 hrs.
Note this online demo is using the current main branch version.

Sample images can be found here

Installation

# First, you need to have Python(>=3.8) installed on your system
# The latest version often does not work with pytorch yet
$ python --version
Python 3.10.6

# Clone this repo
$ git clone https://github.com/zyddnys/manga-image-translator.git

# Install the dependencies
$ pip install -r requirements.txt

$ pip install git+https://github.com/lucasb-eyer/pydensecrf.git

The models will be downloaded into ./models at runtime.

If you are on windows

Some pip dependencies will not compile without Microsoft C++ Build Tools (See ).

To use cuda on windows install the correct pytorch version as instructed on https://pytorch.org/.
Add --upgrade --force-reinstall to the pip command to overwrite the currently installed version.

If you have trouble installing pydensecrf with the command above you can download the pre-compiled wheels from https://www.lfd.uci.edu/~gohlke/pythonlibs/#_pydensecrf according to your python version and install it with pip.

Usage

Demo mode (default)

# saves singular image into /result folder for demonstration purposes
# `--use-cuda` is optional, if you have a compatible NVIDIA GPU, you can use it.
# use `--use-cuda-limited` to defer vram expensive language translations to the cpu
# use `--inpainter=none` to disable inpainting.
# use `--translator=<translator>` to specify a translator.
# use `--translator=none` if you only want to use inpainting (blank bubbles)
# use `--target-lang <language_code>` to specify a target language.
# replace <path_to_image_file> with the path to the image file.
$ python -m manga_translator -v --use-cuda --translator=google -l ENG -i <path_to_image_file>
# result can be found in `result/`.

Batch mode

# same options as above.
# use `--mode batch` to enable batch translation.
# replace <path_to_image_folder> with the path to the image folder.
$ python -m manga_translator -v --mode batch --use-cuda --translator=google -l ENG -i <path_to_image_folder>
# results can be found in `<path_to_image_folder>-translated/`.

Web Mode

# same options as above.
# use `--mode web` to start a web server.
$ python -m manga_translator -v --mode web --use-cuda
# the demo will be serving on http://127.0.0.1:5003

Manual translation

Manual translation replaces machine translation with human translators. Basic manual translation demo can be found at http://127.0.0.1:5003/manual when using web mode.

API

Two modes of translation service are provided by the demo: synchronous mode and asynchronous mode.
In synchronous mode your HTTP POST request will finish once the translation task is finished.
In asynchronous mode your HTTP POST request will respond with a task_id immediately, you can use this task_id to poll for translation task state.

Synchronous mode

POST a form request with form data file:<content-of-image> to http://127.0.0.1:5003/run
Wait for response
Use the resultant task_id to find translation result in result/ directory, e.g. using Nginx to expose result/

Asynchronous mode

POST a form request with form data file:<content-of-image> to http://127.0.0.1:5003/submit
Acquire translation task_id
Poll for translation task state by posting JSON {"taskid": <task-id>} to http://127.0.0.1:5003/task-state
Translation is finished when the resultant state is either finished, error or error-lang
Find translation result in result/ directory, e.g. using Nginx to expose result/

Manual translation

POST a form request with form data file:<content-of-image> to http://127.0.0.1:5003/manual-translate and wait for response.

You will obtain a JSON response like this:

{
  "task_id": "12c779c9431f954971cae720eb104499",
  "status": "pending",
  "trans_result": [
    {
      "s": "☆上司来ちゃった……",
      "t": ""
    }
  ]
}

Fill in translated texts:

{
  "task_id": "12c779c9431f954971cae720eb104499",
  "status": "pending",
  "trans_result": [
    {
      "s": "☆上司来ちゃった……",
      "t": "☆Boss is here..."
    }
  ]
}

Post translated JSON to http://127.0.0.1:5003/post-translation-result and wait for response.
Then you can find the translation result in result/ directory, e.g. using Nginx to expose result/.

Translators Reference

Name	API Key	Offline	Docker	Note
google			✔️
youdao	✔️		✔️	Requires `YOUDAO_APP_KEY` and `YOUDAO_SECRET_KEY`
baidu	✔️		✔️	Requires `BAIDU_APP_ID` and `BAIDU_SECRET_KEY`
deepl	✔️		✔️	Requires `DEEPL_AUTH_KEY`
papago			✔️
offline		✔️	✔️	Chooses most suitable offline translator for language
sugoi		✔️	✔️	Sugoi V4.0 Models (recommended for JPN->ENG)
jparacrawl		✔️		Supports JPN,ENG
jparacrawl_big		✔️	✔️
m2m100		✔️	✔️	Supports every language
m2m100_big		✔️
none		✔️	✔️	Translate to empty texts
original		✔️	✔️	Keep original texts

API Key: Whether the translator requires an API key to be set as environment variable.
Offline: Whether the translator can be used offline.
Docker: Whether the translator is available in the docker image.

Language Code Reference

Used by the --target-lang or -l argument.

CHS: Chinese (Simplified)
CHT: Chinese (Traditional)
CSY: Czech
NLD: Dutch
ENG: English
FRA: French
DEU: German
HUN: Hungarian
ITA: Italian
JPN: Japanese
KOR: Korean
PLK: Polish
PTB: Portuguese (Brazil)
ROM: Romanian
RUS: Russian
ESP: Spanish
TRK: Turkish
UKR: Ukrainian
VIN: Vietnames

Docker

Requirements:

Docker (version 19.03+ required for CUDA / GPU accelaration)
Docker Compose (Optional if you want to use files in the demo/doc folder)
Nvidia Container Runtime (Optional if you want to use CUDA)

This project has docker support under zyddnys/manga-image-translator:main image. This docker image contains all required dependencies / models for the project. It should be noted that this image is fairly large (~ 15GB).

Hosting the web server

The web server can be hosted using (For CPU)

docker run -p 5003:5003 -v result:/app/result --ipc=host --rm zyddnys/manga-image-translator:main -l ENG --manga2eng -v --mode web --host=0.0.0.0 --port=5003

docker-compose -f demo/doc/docker-compose-web-with-cpu.yml up

depending on which you prefer. The web server should start on port 5003 and images should become in the /result folder.

Using as CLI

To use docker with the CLI (I.e in batch mode)

docker run -v <targetFolder>:/app/<targetFolder> -v <targetFolder>-translated:/app/<targetFolder>-translated  --ipc=host --rm zyddnys/manga-image-translator:main --mode=batch -i=/app/<targetFolder> <cli flags>

Note: In the event you need to reference files on your host machine you will need to mount the associated files as volumes into the /app folder inside the container. Paths for the CLI will need to be the internal docker path /app/... instead of the paths on your host machine

Setting Translation Secrets

Some translation services require API keys to function to set these pass them as env vars into the docker container. For example:

docker run --env="DEEPL_AUTH_KEY=xxx" --ipc=host --rm zyddnys/manga-image-translator:main <cli flags>

Using with Nvida GPU

To use with a supported GPU please first read the initial Docker section. There are some special dependencies you will need to use

To run the container with the following flags set:

docker run ... --gpus=all ... zyddnys/manga-image-translator:main ... --use-cuda

Or (For the web server + GPU)

docker-compose -f demo/doc/docker-compose-web-with-gpu.yml up

Building locally

To build the docker image locally you can run (You will require make on your machine)

make build-image

Then to test the built image run

make run-web-server

Next steps

A list of what needs to be done next, you're welcome to contribute.

Use diffusion model based inpainting to achieve near perfect result, but this could be much slower.
IMPORTANT!!!HELP NEEDED!!! The current text rendering engine is barely usable, we need your help to improve text rendering!
Text rendering area is determined by detected text lines, not speech bubbles.
This works for images without speech bubbles, but making it impossible to decide where to put translated English text. I have no idea how to solve this.
Ryota et al. proposed using multimodal machine translation, maybe we can add ViT features for building custom NMT models.
Make this project works for video(rewrite code in C++ and use GPU/other hardware NN accelerator).
Used for detecting hard subtitles in videos, generting ass file and remove them completetly.
~~Mask refinement based using non deep learning algorithms, I am currently testing out CRF based algorithm.~~
~~Angled text region merge is not currently supported~~
Make web page only show translators with API key
Create pip repository