Simple extensible ASS subtitle editor for Linux

Features

• Python - extend it however you want with ease
• Video preview
• Audio preview (spectrogram)
• Audio and video are synced at all times
• Spectrogram range selection for eases timing
• Spectrogram shows where subs start and end
• Slow playback support
• I can sub an entire episode without ever having to touch the mouse
• Mouse users are not excluded and can click their way to all the commands
• Robust plugin API (everything GUI is capable of can be done through the API)
• Simple architecture (Commands ↔ API ↔ GUI)
• Separate control for persistent inline comments (useful for translating)
• Newlines support in the editor
• Seeking is aligned to video frames
• Style editor with realistic preview (taken from video source)
• No bloat

Planned features

• Documentation
  • Hotkeys
  • Built-in commands
  • Configuration files
  • The API
• Meta data editor

Screenshot:

Installation

• Install system dependencies
  • libmpv
  • ffms
  • fftw
  • QT5 and PyQT5 bindings
• Clone the repository: git clone https://github.com/rr-/bubblesub
• Enter its directory: cd bubblesub
• Install bubblesub: pip install --user .

Configuration and plugins

• ~/.config/bubblesub/scripts: contains user plugins
• ~/.config/bubblesub/: contains user configuration in JSON (subject to change, please back it up between upgrades)
• ~/.cache/bubblesub/: used to cache time codes and such

Example plugin: speech recognition of selected lines

import io
import asyncio
import speech_recognition as sr
from bubblesub.api.cmd import PluginCommand


LANGUAGE = 'ja'


async def _work(api, logger, line):
    logger.info('line #{} - analyzing'.format(line.number))
    recognizer = sr.Recognizer()
    try:
        def recognize():
            with io.BytesIO() as handle:
                api.audio.save_wav(handle, line.start, line.end)
                handle.seek(0, io.SEEK_SET)
                with sr.AudioFile(handle) as source:
                    audio = recognizer.record(source)
            return recognizer.recognize_google(audio, language=LANGUAGE)

        # don't clog the UI thread
        note = await asyncio.get_event_loop().run_in_executor(None, recognize)

    except sr.UnknownValueError:
        logger.warn('line #{}: not recognized'.format(line.number))
    except sr.RequestError as ex:
        logger.error('line #{}: error ({})'.format(line.number, ex))
    else:
        logger.info('line #{}: OK'.format(line.number))
        if line.note:
            line.note = line.note + r'\N' + note
        else:
            line.note = note


class SpeechRecognitionCommand(PluginCommand):
    name = 'grid/speech-recognition'
    menu_name = 'Speech recognition'

    def enabled(self):
        return self.api.subs.has_selection and self.api.audio.has_audio_source

    async def run(self):
        for line in self.api.subs.selected_lines:
            await _work(self.api, self, line)

Questions

1. Why not aegisub?

   As of August 2017:

   • It's dead
     • Forum is down for two months
     • Bug tracker is down for much longer
     • Pull requests are ignored
   • It has numerous unfixed bugs
     • It often crashes
     • Recent menu wasn't working for at least 2 years
     • Karaoke joining never worked to begin with
   • Its API is very limited and I can't write the plugins I need
   • Its code is complicated making it hard to add features to

2. Windows builds?

   This is a hobby project and wrestling with Windows to have it compile a single C dependency library isn't my idea of a well-spent afternoon. It should be possible with MSYS2.

Contributing

1. I want to report a bug.

   Please use GitHub issues.

2. I want a feature.

   Chances are I'm too busy to work on features I don't personally need, so pull requests are strongly encouraged.