Warning: I’m a newbie at this, so apologies in advance for things flying over my head.
I mean to transcribe a podcast episode (not my own) that has two speakers, one of whom has audio that’s very muffled. If possible, I’d like to sharpen the audio of just that speaker, but I’ve found myself pretty lost. I’ve just gone through the wiki and the manual and (newbie alert) I’m having trouble figuring out where I should start. There’s the occasional static in that problem speaker, but my priority is sharpening that speaker’s audio. Which effect is best for that?
Other notes: I’m on a Windows 8.1 and using Audacity 2.1.1 (which I’d installed via the .exe file).
Are the two voices on separate Audacity tracks, or do you just have a single file with the two voices mixed together?
If you don’t have the voices recorded separately, then the job is pretty tough. You cannot process one voice and not the other if the voices occur at the same time.
If the voices alternate from one person speaking to the other person, then the first task is to separate the voices by duplicating the track and removing (silencing) one voice from one track and the other voice from the other track, This is a laborious process as you need to select and silence each part manually. (“Ctrl + L” to silence the selected audio).
If you have the voices on separate Audacity tracks, you can use the Equalization effect to make the dull track brighter.
Those words in a post almost always means someone trying to record both sides of a cellphone or Skype call with the wrong software.
Do you know which?
As above, if it’s a mixed track, you’re pretty much dead. You don’t actually have both tracks, you have one track and the echo cancellation errors of the other, and it’s not just muffled, there are actually parts of pitch and tone character missing.
steve: The episode’s in one mp3 file (it’s an interview). I’m afraid I don’t know what you mean by two separate tracks - you mean two separate files?
Kozikowski: I don’t know which.
Okay, thanks. When you select and silence each part manually, do you mean each time they speak? Or one “round,” if you will, of the interviewer asking a question and the interviewee answering?
The file has only one panel on the left hand, so I guess that’s one track only. …Oh well. Correct me if I’m wrong here, but from what I could gather in the manual (I have absolutely zero experience and a lot flew over my head) and a google search, Equalizer or Normalize is the effect I’m using, right?
Equalization is a very, very fancy tone control. Normalize and Amplify (cousins on their mother’s side) are volume controls. They just make things louder or softer.
The real problem is splitting the quiet voice from the loud one so you can fix one without messing up the other. There is no good way to do that. If everything is on one horizontal blue wave, then both of the voices are smashed into one track.
We can’t tell you it’s impossible to split them, but it’s highly unlikely. You have to go into the blue waves with the magnifier tools and drag-select each word or sentence of the far side conversation and apply effects, corrections and filtering. Then select the next word or sentence. It can’t take you any more than a month or so on a long interview.
That’s assuming there’s enough voice quality to rescue. People ask us all the time to “Clean Up” their voice track. The words “Clean Up” are usually the kiss of death. Audacity has no forensics tools. We can’t bring somebody’s voice back to life.
Can you post a short segment on the forum showing both voices?
Oh, that’s different. We still can’t fix it, but it’s different. That’s what happens to a Skype interview when either the connection is ratty, or, the people on both ends are not wearing headphones.
Once the sound does what you have it’s all over. There is no filter to fix that. Wine-glass, honking and bubbling is permanent. The volume difference isn’t all that great. You could almost do nothing and get away with it if the quality was higher. That’s almost the same volume difference between Denise and me. She sounds like she’s sitting on a sofa behind me. She’s actually sitting in her sewing room four time zones away.
