So part way through last week I was frustrated trying to find a way to get a cleaner copy of the audio recording from Skype. It always sounded weird that I always sounded ultra clear and everyone sounded kinda odd due to the way sky is recorded. We use a program called Callgraph to take care of the recording. It records my voice which is local to the left audio channel and everyone else to the right audio channel. I end up splitting these into mono channels so its easier to listen to.
One of the problems that exists is audio noise reduction and equalization is a bit messy from the Skype recording. I usually end up running the file through Levelator and and run noise reduction through Audacity. The end result isn’t the greatest but it is usable. Sometimes you get some very robotic sounds due to the process though.
So this past recording (Episode 17), I asked the team to try something out for me. Since Callgraph records the local channel to the left audio, wouldn’t it be easier to just grab everyone’s left channel and build the audio file with that? In theory it would work just like doing it live assigning each mic to their own audio track in the program. In practice though I learned (or in some cases relearned) the problems with recording audio (or even video) on a PC.
So once we finished the podcast I started getting the files from our hosts through Dropbox. All looked good until I dropped the audio files into Sony Acid Music Studio. I immediately noticed a problem once I synced up My audio track with Calaera’s audio track. Although the first 20 seconds were synced things started to go just a slight bit out of phase until you jump a good half our into the audio where it was off by almost a full second or two! I was about ready to curse when I saw this. Was it the compression she used? No it couldn’t be that. I loaded up Orophen’s audio file and the same problem. It only dawned on me the next day what had happened.
With recording audio (and video) on a PC, the way the audio gets recorded is dictated by a few factors. Audio latency of the hardware, CPU, and HDD read/write speeds. Any background programs such as web browsers and such can effect the recording by a millisecond or two since it draws resources. In most cases it’s unnoticeable however in the way we were attempting this recording, it slowly became very apparent. The same issue comes from video as well as I remember when I tried to record gameplay footage from my console for youtube videos.
I’m a bit stubborn and didn’t just fall back to the old method and instead focused on how to make this work. I started with my recording as a baseline importing my split audio track as well as the original channel with both left and right channels not separated. Since my voice would be synced with the unsplit track I was able to make the cuts for beginning and end relatively easily. I then loaded each host one by one and synced up the start of the podcast. I would then advance through the podcast slowly trying to find a point where our host Calaera’s voice would go out of phase beyond a half a second. I would make the cut and move both the person’s voice and the full Skype recording and sync it back up by moving it either forward to catch up or back to pull it back. Yeah the audio would speed up or slow down depending on how the thing was recording :O
This took a good 3 hours but in the end I had a very clear sounding podcast compared to previous attempts. I was able to apply more directed noise reduction a specific host rather then the whole podcast which some hosts didn’t need. I didn’t mess with a lot of EQ as at this point I was tired of it all and wanted to go to sleep.
In the end after all the work produced one of the most clearest sounding podcasts we have to date. Will I do this again? I dunno, I’m not sure the headaches of editing this is worth it. There is also the time needed to upload the files to Dropbox and then get the link. I’m lucky the new program I had gotten works well this with kind of editing but man, i can’t imagine what would happen if the audio had slowed down significantly on someone’s recording or if someone’s recording fails. Too many points of failure for my taste.
It does give me a chance to play with localized tracks for each host which is something I’ll be facing in a future podcast should my friend ever get it up and running.