How over 200 voices sang as one virtual choir

by | 25 May 2020 | Advancing Barbershop | 0 comments

“Tomorrow’s another day.
I hope and pray we’ll be together.”

As members of Sweet Adelines International, we all know this song by heart. It’s a unifying song that you can sing whenever and wherever there’s a fellow Sweet Adeline.

When the 2020 regional contests and conventions were cancelled in March due to COVID-19, many of us were gutted. A fellow member of my chorus (Jenny Lycett), London City Singers, suggested:

“… all R31 choruses/quartets did a video of themselves singing How We Sang Today in all different settings – could be at rehearsal, in full stage costume, on location, small groups, whatever – and we made a collage of them to release on social media on what would have been contest Saturday.”

I thought, yeah, so I forwarded this suggestion immediately to the Marcomms Team of the Regional Management Team. When Emma Riley, then Region 31 Communications Coordinator, added on with “could we do some sort of virtual choir thing?” along with suggesting a virtual convention, we started to put things in motion.

In this article, I’m going to share with you how this was achieved and some learnings I picked up along the way. Hopefully it will be of use especially if you’re planning to do one yourself.

Establishing guide tracks

As mentioned, every Sweet Adeline knows their own part (or multiple parts) of How We Sang Today BUT every Sweet Adeline also sings it differently – tempo, phrasing and even pitch! Yaiks! I’m no audio engineer but that would be a lot of work to put together, not to mention auto-tuning, to make one cohesive recording, which we weren’t really prepared to do. Emma volunteered her chorus, Vocal Dimension, to produce the guide tracks. Once ready, the tracks were made available to listen to or download from the Virtual Convention webpage.

The guide tracks were simple mechanisms to help the lining-up of the audio easier and faster. This is why we asked people to sing with them even while they record themselves singing their part – not audible, of course – although there were some recordings that had them loud with their own singing and some didn’t even use them at all!

Gathering all submissions

One question I got asked repeatedly was how to send their video recording. We did instruct people to send them by email but video files are usually big, especially if they were done at high quality and resolution. So I told people to send them using wetransfer.com if the file is too big for email attachments. This worked well but please be aware that download links on wetransfer expire so if the intended recipient didn’t download them in time, they won’t be able to get your file at all.

In hindsight, it would have been easier and better if we just asked people to upload their recordings to a cloud file server like Dropbox and email the share link. This way, the person collecting all the submissions is not under pressure of saving them all before the wetransfer link expires.

But we got all the recordings. Well, we believe we did. If we did then you would’ve made it onto the video. If we didn’t, my apologies.

Oh, and if you’re doing this yourself and you’re the intended recipient of the recordings, make sure you have enough space on your computer to store the files. I have a Pro account on Dropbox (max 3TB) plus my laptop has a capacity of 500GB in disk space, which accommodated all the submissions (which was a whopping total of 31GB).

My tools

Before I get into the actual process of how the final video was produced, let me just give you a rundown of the tools I have and used.

Hardware
MacBook Pro 13-inch 2017, running on macOS Mojave

Software
Adobe Audition for audio mixing
Adobe Premiere Pro and After Effects for video editing

All the software is paid as I have an Adobe Creative Suite subscription for my trade (I’m a digital producer/creator/consultant/marketer/designer). Many people say that the same can be achieved by using free tools like Audacity (for audio) and iMovie or Windows Movie Maker (for video editing) but you’ll have to find those tutorials online yourself and find the equivalent of the process I will be describing below.

Creating one mega mix

First step in producing the final product was to mix all the audio together. Why? To make the final video sound good and look seamless. You’ll end up losing a lot of footage if you try and align everything in your video editing software without detaching the audio.

I started working on lining-up the audio even before the submission deadline. I first worked on the tenor part as there were only 22 – easy and manageable, I thought. Using Adobe Audition, I lined up 22 tenor tracks in no time.

A timelapse of the lining-up of the tenors

Then, I began working on the lead part – but as there were quite a number submitted, I decided to group the leads into three sessions on Audition. This was purely for easily managing the process in my digital audio workstation. As soon as I finished one batch of leads, I got so excited and combined it together with a mixdown of the tenor part. Whoa! They weren’t aligned.

It was at this point that I created a mixdown of a quartet first and used that to align every single track submitted. Yes, I listened to every submission not just one time but several times to make sure that everyone is singing at the same time. It wasn’t perfect, and if this was an actual performance before Sweet Adelines judges, it will get a lot of sync comments. I just made sure that everyone comes in at the same time for each bar in the music or key passages. I also tried to minimise the “tttttttomorrow” or “sssssang” or “thissssss” and the loud breathing in between phrases and words.

From a personal standpoint, hearing the audible breaths, and seeing them on the waveform, made me conscious of my own breathing technique as a singer. You might want to try using the TE Tuner app in analysis mode showing the waveform to see if your breathing creates a wave instead of just a flat line.

Back to the mega mix, I finished lining up all the audio submitted after 30 hours. At this point, I had the following mixdowns:

1 tenor
3 leads
3 baritones
4 basses

I added several effects like denoise, parametric equalizer, and mastering to the master track of each of the sessions. I was so excited to put them all in one multitrack session but when I hit play, it sounded horrible! Like everyone was singing into a tin can. Nooooo!

Mega panic set at this point so I consulted with some people who I know mix audio. They suggested things like cleaning up each track – that means going back to each of the 211 individual tracks, and not applying or minimising any compression on the mixdown. After listening to a lot of sage advice, here’s what I did.

  • I went back to every session and removed the effects before creating each mixdown.
  • I combined all voice parts into one mixdown first – cleaning the audio mix for each part.
  • Once I had a mixdown for each part, I combined them all together and only applied the multiband compressor effect to produce the final audio.

Assembling the video

When producing a video, you should know where the final product will be published first and foremost. Why? This will help you establish the size and resolution of the final video for optimal experience for the viewer. For example, if you intend to post a video on your Instagram feed, then the video is better in square (1:1 aspect ratio) and should have a maximum duration of 60 seconds.

For this project, we decided that the final video will be published on Facebook and YouTube so this means it will be a 16:9 (widescreen) and at least 1080p (HD).

Knowing that I have a canvas of 1920×1080, I figured that the easiest way to put the submitted videos in a grid with 16:9 aspect ratio, too. Doing it this way means I don’t have to resize the submission as we’ve asked people to record themselves in landscape mode. However, I did resize a number of videos because they weren’t in 16:9 aspect ratio. To do this, I used Adobe After Effects. This process also helped me standardise the frame rate of the clips used in the final video.

Stacking the clips into the 4×4 grid

I also decided that I won’t be placing all 211 in one sequence throughout the entire song because (1) each one will be too small to see, and (2) I didn’t want to risk my laptop dying because I know that doing that will use a lot of processing power. So I figured 16 per sequence (4×4) would do nicely and went on to calculate that there will be 14 sequences for the entire song. Each sequence will be 6.2 seconds each to give everyone equal air time. However, that exact duration didn’t flow well as it was transitioning at awkward moments in the song. I turned to the lyrics and set the duration of each sequence according to the phrases. Just like in singing, there’s a different sequence for every phrase. This way, it felt more organic.

1  –  Tomorrow’s another day
2  –  I hope and pray we’ll be together
3  –  Tomorrow, this today (today)
4  –  Will be a yesterday that’s gone forever
5  –  So take my hand, my friend
6  –  I want to say
7  –  I’m glad we laughed and loved and sang
8  –  Together today (today)
9  –  So take my hand (take my hand)
10  –  I want to say (I want to say)
11  -  I’m glad we laughed, I’m glad we loved
12  –  I’m glad we sang (oh how we sang)
13  -  Today
14  –  How we sang today

I did attempt to put everyone in one sequence for the last ‘today’ but as suspected, Adobe Premiere Pro started to crash. After checking in with the Marcomms Team to say that it can’t will be difficult to achieve given my current resources, Emma suggested using stills instead. Following another sage advice, I combined and created a mosaic of stills from the sequences for the final shot. It wasn’t ideal but it was the best I could do given the circumstances.

Attempt to put all videos into one

After one last render, the final video was ready to premiere on the region’s Facebook page and YouTube channel in time for the Mass Sing of the Virtual Convention.

In closing

I spent a total of 62 hours producing the final video and learnt plenty along the way. It was a lot of work, repetitive and stressful at times, but after seeing the reactions and comments from many people in the region, it was all worth it.

Would I do something like this again? Maybe. If I do, I probably would be spending less time as now I know what I should and shouldn’t be doing although I’ll also be trying to improve some of the steps in the process.

I wrote another article with a more technical slant on Medium in case you’re really interested in doing one yourself, which I’m hoping will help.