0.6.10

Added

• It is now possible (though usually unnecessary) to supply the source language
  when translating subtitles.

Fixed

• Exported images are now 50% larger in each dimension. Screens got bigger in
  the last 8 years, and 240x160 is feeling a little small.
• Transcribed subtitles now default to 40 characters per line, rather than 30.
  This is more standard, and it uses space more efficently in VLC.
• Improve docs and error messages for export anki. Thank you to @rdearman for
  the feedback!
• Give a more helpful error for overlapping subtitles in input files, which we
  don't really support at the moment (though we can fix small overlaps). Thank
  you to everyone who let me know about this!

0.6.9

Fixed

• Build only substudy and dependencies, to avoid having builds fail due to
  substudy-ui.

0.6.8

Fixed

• dict_importer: Remove build dependency on openssl.

0.6.7

Fixed

• Fixed formatting issue which broken binary releases.
• Improved documentation for transcribe --expected-text.

0.6.6

Added

• Improved help messages for substudy transcribe.
• transcribe now supports --expected-text, which should be used when you
  believe that you have a more-or-less complete and accurate transcription
  already, which you wish to synchronize. This is especially useful for songs,
  where lyrics are readily available but the error rate may be higher. This
  mode also preserves line breaks.
• The transcribe --example-text option is now --related-text, in order to
  make the meaning clearer. The old option is still supported.
• transcribe now supports an optional --lang argument. This is most useful
  when the language is known, but the available "related text" is non-existant
  or extremely short. Normally, the language is detected automatically.

0.6.5

Added

• Run more transcription and translation requests in parallel. This greatly
  reduces the time needed to work with large media files.
• Cache AI API requests. Calling an AI model is slow and costs money. Making
  the same calls over and over again is a waste of time and money, especially
  when we successfully process 99.5% of a large media file. So now we cache
  recent successful requests. So if you need to re-run a incomplete
  translation, it should be much faster and cheaper. (Cache files are stored
  wherever your OS thinks they should be stored. On Linux, this is
  ~/.cache/substudy.)

0.6.4

Added

• We now support substudy transcribe with a --format=whisper-srt option.
  This outputs an SRT file generated directly by Whisper. This seems to have
  more artifacts than the default --format=srt, but it probably works a lot
  better for languages like Japanese that don't use spaces between words.

0.6.3

Fixed

• Linux: Do not use OpenSSL. It prevents us from building static binaries, and
  generally makes everything else harder.

0.6.2

Added

• Add substudy export anki for use with AnkiConnect. See the README.md file
  for an example.

Fixed

• Correctly transcribe audio files that do not tag language information.

0.6.1

Added

• Audio-only files should now export a single album cover (if available) in
  place of video frames.