Dealing with an editing nightmare - could use some advice

Hey everyone,

First post here! I’m an intermediate user of Audacity (I know enough about audacity to edit and make things sound nice, but not enough that I’d know how to solve this on my own), so I’m hoping you all can help me with the issue I’m having. I’ll be willing to provide any information you folks would need, so feel free to ask.

I am editing a podcast, part of which is a discussion portion with a host and 2 participants. The only files I have to go off of for the discussion portion are:

  1. A track, recorded off a Discord bot called Pawa, containing the voices of all discussion participants, split into 2 parts because the bot could only record about 2 hours of audio at a time (I’ll call the first half “1a” and the second have “1b” on further reference), and

  2. A track containing just the voice of the host. This was recorded through audacity.

We tried to get files of each individual participant, recorded through Audacity, but for various reasons those were lost. One of the participants closed their Audacity without saving, and I think the other just never recorded in the first place. I did my best to coach all of them through how to make sure that they recorded and saved these tracks, but apparently that didn’t take. I tried to get the host to use a bot that recorded all participant’s voice tracks separately, but they used Pawa without notifying or consulting me first. These two tracks are all I’m working with. Re-recording is not going to be an option, as much as I would like it to be.

When I got track #1, I prayed that it would be usable, and unfortunately, it is not. The host cuts out very, very badly every single time they talk. My plan to work around this was to make sure tracks #1 and #2 sync up exactly, slowly go through the discussion, and silence all instances of the host talking on their own (with no cross-talk). It would take a long time, but that’s all I can think to do to salvage the audio.

Unfortunately, tracks #1a and #2 do not sync up exactly through the entirety of the project. They sync up perfectly at the beginning, but by the time #1a ends, it’s about a quarter of a second behind track #2. This is just long enough to be jarring and noticeable if I do what I planned to do above without trying to do some kind of tweak or adjustment. The same thing happens with #1b and #2, but to a lesser degree because #1b is shorter than #1a.

So, how should I go about fixing this in the most efficient way? I really don’t know where to start with this.

Any and all input is appreciated, and I’m sorry if this is a dumb question.

I assume that recording is OK. Let’s call that “track 2”.

The first task is to get the length of 1a to match the equivalent part of track 2.

  1. Import both into Audacity,
  2. Try to find an easily identified peak (or valley) near the end of 1b that you can locate precisely in track 2.
  3. Using the Time Shift Tool, line up the start of track 1a and track 2.
  4. Select as precisely as possible from the start of track 2 to the point that you identified in step 2.
  5. From the “Selection Toolbar”, make a note of the length that you have selected.
  6. Select as precisely as possible from the start of 1a to the point that you identified in step 2.
  7. Open the “Change Speed” effect and enter the time that you noted in step 5 as the “New Length” and apply the effect.
  8. Check that the points identified in step 2 now line up in both tracks (if not, “Undo” and go through the above steps again).
  9. “Undo” the Change Speed effect.
  10. Select All of track 1b.
  11. “Effect menu > Repeat Change Speed” (from the top of the Effect menu). This will apply the Change Speed effect with the same speed ratio to the entire track.

Your tracks will now, hopefully, be synchronized.

It will take some fiddling about to get the best settings, but this is probably possible with the “Auto Duck” effect (see: Auto Duck - Audacity Manual).
Track 1b should be the upper track. Track 2 should be the lower track. Select the upper track only and apply Auto Duck.
The Duck amount should be set to -24 dB.
“Fade” lengths will need to be short (experiment)
“Pause” time will need to be fairly short.
When you find settings that work as well as possible, you will probably want to repeat the effect once or twice to increase the attenuation.

Good luck :wink:

Sounds similar to something that happened to me.

Maybe this (my solution) might help? See end of thread.

https://forum.audacityteam.org/t/2-audio-tracks-same-session-timebase-mismatch-help/55747/8

People see Effect > Change Speed and all the odd settings on top (calculate the percentage of the hoo-haa) and freak out. At the bottom of the control panel is show length. “The show is this length. I need it that length. Call me if you have a problem.”

Do Not use either of the other two tools, Change Tempo or Change Pitch. They don’t do the right jobs and can create sound distortions way more often than Change Speed.

You have a Producer problem. Who is directing activities, hiring, or writing checks? I suspect the two guests have no skin in the game and couldn’t care less if the production goes off well or not. That’s pretty deadly with the Local Microphone Recording technique which is, as you found out, a lot like matching cats.

containing the voices of all discussion participants

Is it really that bad? I can listen to a slightly ratty show as long as there are no distractions (such as repeated quality changes and frantic editing). The audiobook publishers insist absolutely that everything match. This is a decision The Producer makes.

Were you listening to the show during the performance? There is a sneaky-pete way to record the total performance. Put an extra computer into the conference with no microphone and record the speaker. Note this computer can’t have a microphone or be part of the performance in any way. It’s just the recorder.

How long was the performance? You mentioned it went over two hours. How much over? Where is it published so I can listen?

Koz

There are techniques of improving chats. Put everybody on headphones or earbuds.

The first voice has normal room sounds but they are wearing headphones, so her voice is clear. The second voice is completely hands-free. She looks swell, but she has Zoom Voice.


Koz

The host cuts out very, very badly every single time they talk.

The Headphone Thing can help with the voice switching business. The chat system never hears recursive room echoes (receive voice bouncing from the walls) and never has to switch voices on and off.

Koz