Pony Sorter

A tool used by the Pony Preservation Project. If you don't know what it does, you probably don't need it.

Getting started

• Clone this repo. Optionally, Windows users can download the binary release.
• Download all audio sources for the episodes you plan to work with.
• Download the label files from Clipper's MEGA, and copy them to /labels.
• Run pip3 install -r requirements.txt, followed by python3 ponysorter_gui.py. (Or download the Windows release)
• Go to Edit -> Add audio path(s). Paste the path or paths where the audio files are stored.
• Wait ~5 minutes for the files to get hashed. This is only done once for new files, and ensures that everyone is working with the same audio sources.
• Go to Edit -> Load episode. You should see a list of all found episodes.

How to use

• Load an episode.
• Listen to all three versions of the line. You can either click the buttons at the top, or use the numeric keys as indicated.
• Adjust the label data if it's wrong.
• Select the audio source that strikes the best balance between quality and noise reduction.
• If you made a mistake, use the leftmost button to go back a line.
• When all lines have been reviewed, go to Edit -> Save changes (or press Ctrl+S) to save your changes.
• You can also save your progress in the middle of an episode, and return to it later.
• The files are stored in /saved_changes. Upload the JSON files for completed episodes, and post them to the thread.

Export options

Split by pony

Automatically splits audio samples from all episodes for a selected pony, and generates metadata in LJ Speech format. This takes about 5 minutes. Options are hardcoded for now: Twilight only, defaulting to iZotope for unreviewed lines.

Generate combined source

Not yet implemented. Will create a single Frankenstein file for each episode with the cleanest audio source for each line. Potentially useful for existing splitting scripts.

Generate Audacity labels (current episode)

Exports the episode data loaded into memory to /exported_labels. Four files are created. Files suffixed "original", "izo" and "unmix" contain only the lines chosen for each source. The file without a suffix contains all lines.

Generate Audacity labels (all episodes)

Same as above, but exports labels for all episodes. Save your changes before running this.