Sunday 18 July 2010

From Poverty to Africa: Retrospection Part 4 - Audio

Amongst film-makers it is often said that audio is half of video, which is, of course, nonsensical. What they mean is that audio is just as important as video, and here I agree whole-heartedly. If the sound is bad, it doesn't matter how good the visuals are - it's ruined. Making the sound more convincing, adding better sound effects and an appropriate, good quality musical score all make a massive contribution to the film's overall impact. Do not compromise here. The sound is extremely important.

Which is yet another reason why From Poverty to Africa is something of a failure. Cheap, consumer-level camcorders with their built-in microphones are simply not good enough. We realised this quite early on. There was no doubt: we needed an external microphone, so we got a Røde VideoMic, a fairly inexpensive hypercardioid shotgun microphone that had good reviews. And indeed, it was very effective. Compare, for example, scene 18 (in the Tent Modern), which was filmed without the microphone, with scene 9 (the wedding reception) or any of the green screen scenes. Nevertheless, a whole raft of new issues came surging to prominence as a result of the use of this microphone.

Firstly, the Røde VideoMic, like practically any film accessory, is mounted on the camera's standard hot shoe mount. (If you've never heard the phrase 'hot shoe mount' before, do pause for a while to be amused at how ridiculous the name is.) But guess what? Our useless, cheapy consumer-grade miniDV camcorders were so useless, cheapy and consumer-grade that they did not have the hot shoe. Not even a cold shoe. Not even a tepid clog. Furthermore, only one of our three cameras had a microphone port. At first, we simply fixed the microphone to the camera using elastic bands, which worked, but was rather clunky. We also had an accessory for the microphone called a 'dead cat', which is a wind shield that you fit over the microphone. (Pause: you put the microphone on a hot shoe and then put a dead cat on it.) With the microphone fixed to the camera by means of elastic bands, the distance between the microphone and the lens was not great, and so the fibres of the dead cat were occasionally visible in the shot, flapping in the wind. Observe this phenomenon, for example, in scene 8 (at Notre Dame) - though I managed to counter it somewhat by adding the vignette effect. Eventually we got a new camera that had both a microphone port and a hot shoe. Unfortunately this doesn't solve the 'flapping dead cat' problem entirely, so it's worth watching out for that if you use this accessory yourself.

Furthermore, there were a few scenes in which the crew forgot to turn the microphone on. This may seem like a problem for which there is no solution (except re-taking), but remember that we did have multiple cameras in almost all scenes, so there was still the crappy sound from the other camera's internal microphone. In most cases, however, this sound was intolerably bad, and we had to resort to the dubbing and narration solutions. The latter is much easier and more effective. The idea is that Thomas and Questa are narrating the story in the future, so for any scene that did not have usable audio, we could simply record a 'narrated' version of it and make the clip's 'native' audio very quiet. For narrative reasons, however, some parts had to be dubbed.

Dubbing


Dubbing is not easy. Seriously, you do not want to get involved with dubbing. If the actor's lips are visible, you need to get the actor to record their lines such that they match their lips exactly. This takes a few tries, but does work. If you're dubbing a long scene, though, it'll take ages. Thankfully we only had a few isolated lines that demanded dubbing: those in scene 17 outside the Tent Modern (from "I wonder where we should start") and those in scene 39 in the restaurant (such as "Thanks for your advice and recommendation..."). A further problem is that dubbed audio will not sound as though it was recorded in the location where the video was recorded. This is less noticeable if the whole scene is dubbed, but if the scene consists of some dubbed and some non-dubbed sections, it is surprisingly noticeable and jarring. This is an excellent example of how audio is so unconsciously important. We don't generally notice that our voices sound different if we're in a car park compared to if we're in a living room or a big hall, but record yourself saying the same thing in two different locations and you'll notice a marked acoustic difference. And when you're dubbing, the actor is recording straight into the microphone; on location, the microphone is further away (unless it's on a boom, which it wasn't in our case). Making a recording sound like it was recorded on location instead of in a studio (or, in our case, in my bedroom) is very difficult, although tools do exist for this purpose.

Then there's the problem of ambient sound. In a restaurant, you can hear the sounds of other diners; in London, you can hear the traffic noise; and just generally, in any environment, there is some background noise. Even in an empty room with nothing going on, there is some kind of 'room tone'. If you dub some speech, then the parts of the scene in between the speaking will be completely silent, and this is actually very jarring and unnatural-sounding. You need appropriate ambient noise. The best thing is to make a point of recording a small amount of 'nothing' at every location, just in case you run into situations like this: this is called a 'wild track'. Thankfully, we had plenty of ambient sound even when we hadn't specifically recorded a wild track. (This was, in fact, one of the things that we had planned for in advance.)

A few scenes were intended to be dubbed all along - these were the scenes that took place in the car while it was moving. It was rightly deemed that it would be too unsafe to act in a scene while driving on the motorway. What we (they) did was to record the actors mouthing their lines (or more accurately, just mouthing anything) while the car was moving and then record the lines themselves separately. In other words, there was no attempt at lip-synching here. This makes scenes 31 ("This car smells quite fresh...") and 45 (the start of 'Race to the Finish') somewhat unconvincing.

Like I say, stay away from dubbing at all costs. If it turns out that your sound is bad, don't think, "Oh, we'll just dub it!" Seriously consider filming the scene again. Check the sound quality thoroughly during the shoot to prevent issues.

Noise Removal


If you use a camera's internal microphone (don't, though), you will always pick up the hum of the camera's electrical magic. Such a regular and consistent hum is an easy target for automated noise removal, such as can be achieved with Audacity. I used Soundtrack Pro, however, and achieved good results. The first advantage of having an external microphone is that it gets the microphone away from the camera and so lessens the hum noise. This is a rule: get the microphone away from the camera and as close to the talent as possible. For professionals, that means using a boom or clip-on microphones. Seriously consider these options. We didn't, but I definitely believe that it would have had a massive positive impact on the film's overall quality. Even if you can't go to this length, I cannot stress enough the importance of getting the microphone as close as possible to the actors. Look at the beginning of scene 42 (Loch Ness): you wonder why we subtitled it? The fact that Emma (played by Starbuck Friend) was speaking Scots was just a convenient excuse to make the subtitles obligatory, but the real problem is that you can barely hear her. Yes, there's a lot of traffic rocketing past the shores of Loch Ness; yes, it was windy; yes, there was someone playing the bagpipes not far off; but if the microphone were closer to her, it would have made a big difference.

Scene 40 (outside the George Inn in Frocester) is another scene spoilt by traffic noise. In this scene we were successfully using the Røde microphone and we placed it as close as possible to the actors. But we were still filming right next to a road, and despite the fact that Frocester is a tiny place in the 'nowhere' region of the English countryside, cars are still noisy when they pass through it. It is impossible to remove traffic noise in post-production. You can try, and you can let me know if you have any success, but as far as I'm concerned, it's impossible. Noise removal works best when the noise is a regular sound at a particular frequency, or when it's at a particularly low volume relative to the sound you want; in the former case you can use equaliser filters, high cuts and low cuts to make sounds at particular frequencies quieter, and in the latter case you can use a gate to cut off anything in a particular frequency that's below a certain amplitude threshold. This works wonders for electronic hums. But the noise of a car going past is at practically every frequency and is very loud. I'm sorry, but if a car goes right past you during a line, you need to re-take. If it happens too often, you need to go somewhere else, or get clip-on microphones. You don't want to resort to dubbing, after all.

Soundtrack Pro was also useful for certain effects, such as the 'megaphone' effect on the ferry and the 'loudspeaker' effect on Claire's voice when she's on the telephone, both achieved with low pass and bandpass filters, as well as a compressor and reverberation. The stock sound effects that come with the program were also invaluable, and added a lot of value to the sinking ferry and the exploding hypermarket (for example).

Advice


Technical audio-related advice:
  • Before filming, consider all the sources of potential noise. If possible, purge them. If the environment is too noisy, think again.
  • Use an external microphone.
  • Get the microphone as close to the talent as possible. If possible, use a boom or clip-on microphone.
  • Don't forget to turn the microphone on!
  • Point the microphone in such a position so as to maximise the pick-up. Our microphone was hypercardioid, which means it picks up mostly from a region right in front of it and a smaller region right behind it (i.e. not so much from the sides).
  • Record a wild track at each location - i.e. a small amount of ambient sound. It can be surprisingly useful even in non-dubbing situations.
Next we'll consider the music.

No comments:

Post a Comment