Behind the Scenes of MY sigRA: My Audio Editing Workflow

As promised, here is a snippet of the (you may be surprised) hard work that goes into making medical education videos. And we’ll start off with just audio.

My technique is certainly not perfect, and every project teaches me something new. Just producing this video led to me having to solve multiple problems requiring multiple takes: lighting issues, noise, my own uhm’s and ah’s, and many more. Since it’s not “medical education”, I’m publishing it on my personal channel instead.

Don’t forget to subscribe and share!

Virtual Live Anaesthesia Workshop - Regional Anaesthesia for the Chest

If you came here just to watch the video, here you go:

As usual, production notes follow.

The company that handled the live webinar definitely delivered, allowing footage from 3 different areas in Malaysia to be broadcast live via YouTube with only insignificant hitches. There was only 1 pre-recorded session; everything else was live. There was no noise to fix (thank God!). For me there were only a few things to do/fix:

1. To open the video, I created a small clip of the flyer panning the profile of the speakers and schedule, which then transitions into the introduction.

2. In a live broadcast, there would obviously be breaks between sessions. Even a few seconds is unnecessary in this on demand viewing format, so I deleted most of the unnecessary pauses (and even some dialogue) and added some simple transitions.

3. The speakers all speak obviously in different volumes (even in the same clip), so these were fixed with a compressor plugin, with adjustments made only to the input gain for the different speakers (and even with the same speaker but different segment. The aim is based as my now usual norm as per the recommendations here: around -12 to -15; there is no risk of peaking/clipping, but the listener will have to turn up the volume a bit (but avoid distortion). It was too tedious to fix every segment where levels drifted within the same clip; only those obvious (too soft dialogue, noises like sneezes) were edited. I noticed that my job wasn’t perfect, ie there were was some segments peaking at -9 on average, and there was some averaging -15, but in editing an almost 2 hour video as a non-pro (coming out as a 7 GB H.264 file) , I think my job was passable enough and I won’t go for perfect. Best to listen on dedicated speakers away from outside noise.

4. Only minimal highlighting was required (where the somehow the pointer of the ultrasound machine did not appear in the broadcast). These were easily solved with a few arrows and titles.

5. In the Q&A session, where both venues were broadcast simultaneously, I noticed that the audio from Johor lagged behind the video a bit. Wherever it was significant (e.g. where the team from Johor were answering questions), I pushed the audio back (around 17 frames in a 30fps video).

Audio effects used.

The compressor settings

A lil’ bit of low boost for everyone

Fixing the audio lag

Bringing the audio levels down but more even overall

There were of course some stuff that I have no power to fix:

1. There was some echo in the HKL demo, probably from the panelist mics picking up the audio broadcast from the moderator. I don’t know what is the fix for this, unless it is intended for only the speakers to hear from the moderator/other speakers.

2. The footage from Gleneagles could benefit from a bit of color adjustments. But since I’m using a cheap TV as my laptop monitor for now, I can’t fix it.

But overall, I do feel that this is a distinct polishing job of the already very good live footage; this was not a simple copy and upload. Enjoy.

(Do inform if you see any obvious need for improvement of the presentation; otherwise the version uploaded is final)

Regional Anaesthesia Refresher 2020: Back to Basics - Lower Limb Blocks

If you came here just for the video, do take note that the audio is not so good until the 1:13 mark. Here you go:

The audio gets better after minute 1:13. Delivered by Dr Fakhzan. If you enjoy these videos, do show your support by subscribing to this channel.

Read on if you’re interested in the production.

Producing this video made me cry and took a piece of my life. First, the designated audio recorder started a bit late; using the DSLR camera mic is never truly ideal even in the best of times due to the distance from the speaker. But to make things worse, there was a huge amount of noise that I’m not really sure where it’s from (computer/projector fan? Air conditioning?). So the voice on the beginning when it’s all said and done is still not good after editing. Luckily after about 2 minutes the audio is much bettter, but only after removing the noise.

With previous edits, the noise was relatively much easier as it hovered around a single frequency. But not this one:

Pure noise in the first part…

…And the second part from the designated recorder

With the previous videos, the noise hovered around a single frequency, which made editing relatively easy by attenuating the specific narrow frequency range without sacrificing too much of the speaker’s voice. But here, the noise was present throughout the frequency range, whether it was the lows/mids/highs. I tried everything with a combination of EQ, compression, noise gating, denoising to no avail. I was about to give up altogether after spending at least 12 hours of total precious time.

…Until I came across Brusfri, an app by Klevgrand that allegedly can denoise without degrading the original track. There’s 2 versions: $14.99 for iOS and $59.90 for macOS/Windows. To use it, you “select” the portion of the tract with noise and let the app do its magic. I bought the iOS version (which isn’t that cheap over here) that comes with a significant drawback: you can’t really select the range where the noise is. You have to listen to the track then press the learn button to have the app listen to only the part with the noise but not the speech. Once the app has learned the noise pattern, you would notice that the waveform would change and can be played back with noise reduction on or off. You can also do some fine tuning with the noise reduction.

While the track is playing, press and hold the listen “ear” button to have the app analyze the noise…

…Until you release the button

Now you may toggle to hear the track with the noise reduction present/absent…

…Or have it listen again if you think you got the noisy part of the track wrong (noise only with no voice/instruments playing)

A multitudes of options (swipe left or right) to fine tune the noise reduction; adjust by dragging the corresponding line

HPF: High pass filter

This was truly a lifesaver when it came to the humble recording via Voice Memos on an iPhone. From this:

…To this:

Unfortunately it could not greatly improve the inherently not so great voice recording on the beginning of the tract, even with further effects done afterwards.

Before:

And after:

These were the further edits to both tracks after removing the noise:

The first minute

The rest of the video

There was certainly a need for highlighting with this presentation, more so than the one preceding it. To save time, I used simple arrows and highlighting via masking to get the desired effects.

The big lesson here is to get the recordings right the first time, as repairing things in post can be either difficult (or expensive). When pro’s say that listening to bad audio is fatiguing to listen to, they are definitely correct and I can attest to that. Maybe in the future I might explain why as regards to the video, it has simply been mostly still pictures with barely any true video. All in all I probably spent at least 18 hours just to produce this (due to the troubles with the audio) and that’s not productive enough. Hopefully noise is no longer a big problem in the future.