Hey guys, I want to share with you this application that I build to solve one problem that is common on original anime blue-rays - The lack in the native subtitles (Brazilian Portuguese for my case).
The problem
Most media content on the internet is in English, Anime episodes included. I consider myself proficient on English, I can write, read, listening and speak but is not my native language, I’m from Brazil my native language is Portuguese. But, if I can read the Anime subtitle on English — What is the problem with that? — Comfort. I can read the subtitle on English, but my vocabulary is short, and limited by technical English, that is the content that I consume more, so, if some word that I don’t know appears on subtitle, I have to stop the episode, and translate, this break the attention, in my native language is more comfortable and fun to watch.
Ways to solve
There are some ways to solve this problem, the most common is to download subtitle files from the internet, there are many sites that provide subtitle files in many languages, but this approach has some problems:
- Not all animes have subtitles in all languages, some animes are very niche and don’t have subtitles in many languages.
- The quality of the subtitles is not guaranteed, some subtitles are poorly translated, or have many errors.
- The timing of the subtitles may not match the video, causing a bad experience.
- It requires manual effort to find, download and sync the subtitles with the video.
Another way to solve this problem is to use machine translation services, like Google Translate, DeepL, etc. But this approach also has some problems:
- The quality of the translation is not always good, especially for complex sentences or idioms.
- The context of the dialogue may be lost, leading to awkward or incorrect translations.
- It may require internet connection, which is not always available.
The solution: AnimeSubs
To solve this problem, I decided to build an application that uses Large Language Models (LLMs) to translate anime subtitles from any language to any other language that the model supports. The application is called AnimeSubs, and it works as follows:
Requirements: mkvmerge (part of mkvtoolnix) and ffmpeg must be installed and available in the system PATH.
Note: The application is in its early stages, may you encounter some bugs (mainly on Windows and Mac, I do not focus on those platforms right now), this is my first desktop application using Tauri, so, please, report any issues on the GitHub repository.
Download animesubs here: AnimeSubs
- The user provides the video file or folder containing the video files and the LLM provider.
- Select the source and target languages for the translation.
- Select the output type for subtitles (ass, srt or vtt)
- After configuring the settings, the user can start the translation process.
Is very important to select the use mkvmerge for muxing the subtitles into the video file.
Because ffmpeg have some issues with subtitle muxing.
If you do not select mkvmerge, the subtitles may not work.You can add a custom prompt to, for example, if I just keep the default english -> portuguese prompt, the model can mix Portugal portuguese and Brazilian portuguese, the two are correct, but have different nuances, so you may want to specify which one you prefer.
If you want to select some specific subtitle track, select manually.
Click on start translation button and wait for the process to finish (do not close the app on muxing process).
Conclusion
AnimeSubs is a powerful tool for translating anime subtitles using LLMs. It addresses the common issues faced by anime fans who struggle with language barriers and provides a more comfortable viewing experience. By leveraging the capabilities of modern AI, AnimeSubs can deliver high-quality translations that respect the nuances of different languages and cultures.
I hope you find this application useful and enjoy watching anime or other media in your native language! If you have any questions or feedback, feel free to reach out to me on GitHub. Happy watching!
Note: AnimeSubs is licensed under the AGPL-3.0 License.
Top comments (0)