Mirumoji is an open-source, self-hosted Japanese immersion toolkit. Drop in any video, anime episode, drama or audio clip and it gives you: clickable tokenized subtitles with dictionary pop-ups, Whisper-powered transcription, instant SRT/clip extraction, and one-click Anki deck export — all in Docker, all running on your own machine.

Optional OpenAI Integration → Customizable GPT Breakdowns of Subtitles

Optional Modal Integration → Install a CPU only version and run all GPU tasks on the cloud

View a Preview of the frontend (No backend running)

Access full Documentantion

Upload your local anime/J-Drama espisodes or any Japanese video and .SRT subtitles.

Subtitles are tokenized (kuromoji.js), allowing you to click individual words for information

Dictionary (kotobase) page allowing wildcard searches.

Copy and paste text for tokenized output with furigana and clickable words.

Modify System Message and User Prompt with {sentence} and {word} variables.

Upload videos in various formats; they can be converted to MP4 for optimal playback.

Generate subtitles for your videos. Runs FasterWhisper with modified parameters to increase accuracy for longer media such as Anime/ J-Drama episodes.

Transcribe Japanese audio from recordings or uploaded files.

Profile configurations and all other profile-related data is stored and managed via SQLite database by the backend.

Save important video segments with their associated word breakdowns and export as an Anki Deck

Both the frontend and backend have pre-built Docker images and are set up to work with Docker Compose.

The easiest setup is to run the launcher for your platform. For detailed instructions please refer to the Setup Guide

You can also choose to build the images locally with the Dockerfiles provided.

Compressed sizes for the Docker Images