Anime Translation Initiative

AI-helped transcription and translation
Everything works offline

LOAD THE FUNCTIONS

source snippets/enviromentvariables.sh #YOU MUST EDIT THIS ONE
source snippets/functions.sh
source snippets/opus.sh
source snippets/timeformat.sh

Workflow

• I'm assuming you are using linux, and check Dependencies
• More information about each component best researched in their own websites
• The main workflow is as follows:
  • Setup the Model, Whisper here, and use it to translate directly from japanese audio to english text.
  • To get the audio, formated appropriately, you use ffmpeg.
  • Instructions Here
  • The timestamps often aren't completely aligned with the sound, so we can use an AutoSync: ffsubsync.
  • Instructions Here
  • Next comes the human to fix the translation, split long captions, further align them, etc.
  • I propose the usage of Subed, which is an Emacs package
    • Subed allows us to:
    • Watch where are captioning in MPV
    • Efficiently move, merge, split the timestamps, with precision.
    • Instructions Here
  • Then, to fix grammar or spelling mistakes, we can use the Language-Tool
  • Instructions Here
  • Finally, we can load the .vtt file with mpv and enjoy, Instructions Here
• Some extra tools at your disposal:
  • The Opus Model is a text-to-text translator model, like Google-Translate
  • Instructions Here
  • There are two extra tools to align the captions: a Visual Scene Detector(Scene-timestamps), and a Human Voice Detector(Speech Timestamps)
  • Often the captions align with those
  • Instructions Here
  • You can use Whisper to translate a snapshot of what you are hearing from your speakers, using the Speakers-Stream thing
  • Instructions Here

    Setup

    Model Setup

    used model: WHISPER
    will download model ggml-large.bin from: here

    make setup

Model Usage

get audio from video file for whisper to use

useWhisper

the -tr flag activates translation into english, without it transcribes into japanese

Warning

• whisper often breaks with music segments
• if you see it start outputing the same thing over and over, interrupt it
  • then use the -ot milliseconds flag to resume at that point
• After interrupting, copy from Terminal, then format appropriately with:

  formatToVtt

  MPV

  get mpv to load some subs

  mpvLoadSubs

  what subs?
  git clone https://github.com/tekakutli/anime_translation/

VTT efficient creation or edit

I use subed
git clone https://github.com/sachac/subed
add Subed from configAdd.el to Emacs config.el
alternatively, add this extra:
git clone https://gist.github.com/mooseyboots/d9a183795e5704d3f517878703407184
add Subed Extra Section from configAdd.el to Emacs config.el

AutoSync the Subs

This ffsubsync-script first autosyncs japanese captions with japanese audio, and then uses those timestamps to sync english captions to japanese captions.
The japanese captions only need to be phonetically close, which means that we could use a smaller-faster model to get them instead, ggml-small.bin, here.
This is the reason behind the names, why some are called whisper_small vs whisper_large (the model used).

make installffsubsync
autosync

Other Utils

To .srt Conversion

vttToSrt subs.vtt

Export final .mp4 with subtitles

exportSubs

To format a given time in milliseconds or as timestamps, example:

#timeformat.sh has these two commodity functions:
milliformat "2.3" #2 minutes 3 seconds
stampformat "3.2.1" #3 hours 2 minutes 1 second

Grammar-Spelling Checking Language-Tool

Install full-version of Language Tool

make installlanguagetool

Activate it

languagetool

add LanguageTool section from configAdd.el to Emacs config.el
Emacs use:

(langtool-check)

Local Text Translation

your FROM-TO model is either here or here
example, to get the models I use:

make opusInstallExample

edit PATH_TO_SUBS/Opus-MT/services.json appropriately, then:

make installopus

To activate:

#opus.sh has commodity functions
Opus-MT

To use:

t "text to translate"

Get Event Timestamps

Scene-timestamps

Visual Scene timestamps:

make installSceneTimestamps

sceneTimestamps

VAD, Speech timestamps

What is VAD? VAD means: Voice Activity Detector
It gives you the speech timestamps, when human voice is detected
first install torch, then:

speechTimestamps

Translate the Speakers-Stream

you'll need to Ctrl-C to stop recording, after which it will translate the temporal recording

streamtranslate

if you were to have sway, you could put this in your sway config, and have an easy keybinding to translate what you are hearing

bindsym $mod+Shift+return exec alacritty -e bash /home/$USER/files/code/anime_translation/snippets/streamtranslate.sh

Dependencies

• Linux, Bash, Mpv, Ffmpeg, Emacs, Subed
• Whatever model you wish to use
• Python if you use the "Get Event Timestamps" things
  • The vad.py thing downloads silero-vad by itself
• Docker for the LibreGrammar(Language Tool) or the Opus things
• If you want to translate your Speakers, you need pipewire
  • As commodity, you will need: wl-copy and wl-paste, if running on wayland
  • If you don't want them, remove them from streamtranslate.sh

    Why X

• Why Git over Google-Docs or similar?
  • Version Control Systems (git) is an ergonomic tool to pick or disregard from contributions, it enables trully parallel work distribution
• Why .vtt over others?
  • whisper can output vtt or srt
  • subed can work with vtt or srt
  • why vtt over srt? personal choice, but:
  • vtt has no need for numbers as id
  • seems shorter and more efficient