Transcription Aids for Radio Recordings?
by MarjaE from LinuxQuestions.org on (#6NRJX)
I'm trying to transcribe some choppy radio recordings.
I tried doing everything by hand. I have hearing issues, so this was slow and painful going: hear several words, stop, type a few words, back up, hear several more, type a few more, back up...
I tried doing everything with speech recognition software. I most recently tried Speech Note, but Vosk Large crashed, while Deepspeech Mozilla and Coqui were unrecognizable. I think these assume everything is a coherent sentence, without static, interruptions, etc.
There are a couple thing that might help:
1. An app that combines audio controls with a text editor, so it's easier to pause, type, and resume, without having to switch focus.
2. An alternate Speech Note model and/or an other app that transcribes much more phonetically, and doesn't try to turn things into coherent sentences. Bonus points if it can recognize numbers, if it can distinguish voices, even just numbering them, and if it can admit when things are too garbled. I'd be able to go over this with the recording, and fix it, substitute personal and place names, etc.
Any suggestions?
I tried doing everything by hand. I have hearing issues, so this was slow and painful going: hear several words, stop, type a few words, back up, hear several more, type a few more, back up...
I tried doing everything with speech recognition software. I most recently tried Speech Note, but Vosk Large crashed, while Deepspeech Mozilla and Coqui were unrecognizable. I think these assume everything is a coherent sentence, without static, interruptions, etc.
There are a couple thing that might help:
1. An app that combines audio controls with a text editor, so it's easier to pause, type, and resume, without having to switch focus.
2. An alternate Speech Note model and/or an other app that transcribes much more phonetically, and doesn't try to turn things into coherent sentences. Bonus points if it can recognize numbers, if it can distinguish voices, even just numbering them, and if it can admit when things are too garbled. I'd be able to go over this with the recording, and fix it, substitute personal and place names, etc.
Any suggestions?