Making creature vocals using recordings of animals and birds often requires multiple sources, sometimes multiple species, in order to cover the range of emotions needed. For example, in The Revenant I used four species to create the vocalizations and breathing for the bear. At some moments the bear needed to sound ferocious, sometimes it was nurturing with its cubs, and sometimes curious as it investigated the Leonardo DiCaprio character lying on the ground. Then it had to sound in great physical stress after being shot, and finally we needed it to sound like it was dying.
It turned out that there was no way to cover all those bases by using only recordings of bears. Still, everything had to be believably bear-like and consistent. Recordings of bears, dogs, a horse, and my voice did the trick, but required lots of editing and manipulation.
I also tried lots of things that didn’t work.
There are very few recordings of bears sounding terrifyingly ferocious, so I attempted to cheat things like leopards and elephant roars in those cases, but they didn’t sound enough like a bear, and Alejandro Iñárritu, the Director, wanted everything to sound completely believable.
Sometimes you can only find one or two great source vocals that express a given emotion, but you need that kind of vocalization in five or six places, so you have to create variations on the ones you have. Dynamic pitch changing is one approach,but in a case like The Revenant bear I couldn’t pitch anything so much that it would generate unnatural sounding artifacts. Often, just changing the pitch of the beginning or the end of a sound will be enough to make it convincingly different from its source while still sounding like the same creature. Sometimes a dynamic pitch change can be great, but with voices, dynamic pitch changes usually need to be abrupt.
Nearly all real-world mammal and bird vocals will jump from one pitch to another over a short period of time, so gradual pitch changes that you might try will tend to sound unnatural.
Reversing a vocal, or part of a vocal, can also be a way to create a usable variation on an existing recording. But any transients in the sound, like vocal rattles, will usually have a telltale backward sound when they are reversed, so when possible, you’ll want to cut out the transients.
If your source sound doesn’t have a great ending, you can try reversing the beginning of the sound and putting it at the end, though this will sometimes require a bit of reverb to keep the end from sounding too abrupt.
Since there were no recordings of a bear terribly wounded or dying, I had to find ways to fake it. I needed something that would convey the idea that she was in great distress physically. None of the bear breath sounds I had access to worked. I finally found a recording of a horse with severe respiratory problems, gasping for air and vocalizing at the same time. (I’m told the horse survived.) It certainly sounded like a wounded, large animal, but it didn’t sound very much like the bear vocals that had to precede it. The change sounded too abrupt.
My main trick to make the transition work was to begin cheating a few of the horse breathing/vocal sounds into the mix earlier, even before the bear was shot. Like this:
bear vocal, followed by combination of bear breath and horse breath, followed by another bear vocal, followed by another combination but with the horse breath/vocal a little louder than the last time, etc.
That way the audience slowly gets used to hearing this new kind of sound, and isn’t jolted by a sudden change. By the way, I used my own voice earlier in the scene, when the bear is sniffing Leo as he’s lying on the ground. I couldn’t find any ultra- close sniffs from a bear or any other animal that also had a vocal quality, so I recorded myself, and pitch it down about two octaves. I did at least thirty or forty takes before I managed to do a vocal performance that was credible as the bear.
After the bear and Leo have tumbled down the hill, and the bear is breathing its last few breaths, I used a recording of a large bulldog’s wet, glottal breathing. Immediately after we hear that last, weak breath, the baby bear calls in vain for its mother. It’s a heartbreaker.
Any time you need to edit two or more sounds in sequence that you want to feel like one sound, or at least the same kind of sound, the trick is to unify them, to make them similar to each other in as many ways as you can. This also applies to situations when you want two or more sounds to play simultaneously and seem like one sound.
1) spectral unity: choose them or eq them to share as much of the audio spectrum as possible.
2) spatial unity: pan them to the same place, or dynamically with each other
3) dynamic level unity: choose or adjust them so that they follow similar loudness curves
4) dynamic pitch unity: choose them or adjust them so that they follow similar pitch change curves
5) spatial acoustics unity: make sure they have similar reverb characteristics.
As always, set yourself free to experiment and make mistakes. Trying things that don’t work is absolutely necessary in any creative process, even when you’ve been doing it for almost fifty years, like me.