Sep 27, Tue of Week 1 of (post-) Summer 2022
I watched two more episodes of HUNTER x HUNTER (1999), with generated JP sub (for 1.5 episodes), and it now catches up with the recently finished manga volume. So, time to read further again.
I also watched an episode of Yurucamp S1 with JP sub (found in ็ใฃๅญ).
I played around with audio transcribing software for HUNTER x HUNTER (1999), and found openai/whisper to be demanding for a decent transcribing speed. Also, dedicated GPU RAM > 4 GB is required for the recommended medium
and large
models. small
model doesnโt only suck sometimes, but also refused to transcribe for some time segments.
The generated results are of *.vtt
format, and previously used Aegisub doesnโt support this (although I have another script to convert to *.srt
first). Also, Aegisub development was discontinued for like 8 years.
Looking at the content of *.vtt
(with a text editor), and finding that *.vtt
can easily be edited with a text editor or simple text processing scripts; so I want to be able to visualize *.vtt
directly. I found Subtitle Edit, which is actively developed to present time.
The real pro of this app though, is that it can generated subtitles from movies (not only audios, which are indirectly movies). Of course it can generate ones from audio files as well, with timing, and good GUI for proofreading. I also found other audio transcriber softwares, like CLI ones, in Github.
Transcribing quality is sometimes worse than openai/whisper, but at least it doesnโt generate texts for silent segments, or repeated same texts whether (unrelatedly) spoken or not, as well as generated texts are well-timed.