Flies as musicians (8) — Nature-Symphony 73 (improved version — slowed)

(Improved version — see further below)
Nature-Symphony 73 (Flies as musicians 8: Hilltop celebration: freaky sarabandes!) — Surely the most striking yet of my flies-based Nature-Symphonies (and louder than the previous ones), and surely the most hilarious, taking advantage of a further noise-reduction breakthrough (see Techie stuff further below). This has 12 layers, and still rather less background sound and greater clarity than the previous one (9 layers). A really crazy-sounding fun piece, using a gutsy undulating motif built from three potent four-note chords, its included augmented triads giving it a particular sense of depth and seemingly meaningful colour.

Again I retained some aeroplane incursions, indeed rather more than previously. Of particular note is a long episode with at least two propeller aeroplanes rather overlapping, and a quieter jet getting a little into the act too. The chord formed from the motif is clearly audible in the plane sound, with an organ-like quality.

At times grasshoppers figure in the work. Also, the odd bird calls give punctuation here and there. The distant squealy bird calls, I'm pretty sure, are a juvenile buzzard. They often sound to me like a seagull crying out in pain! However, their natural sound is difficult to notice in this rendition, because the top layer is pitched a fourth higher than the original, and again the bottom layer is pitched at an octave plus minor third below original.

Note that the striking way the work draws to an end was NOT at all contrived by me; it's just how things came together. The really strong fly incursion then is remarkably like a close lightning strike!

Improvements in this version
To many listeners the improvements would seem pretty subtle, but there are just two:
(a) The sound is slowed by an amount that reduced the pitch by a minor third — the pitch then being restored. This doesn't affect the musical speed, which is determined by the track offsets, but it does make all individual sounds slightly longer, so that short notes sound more musical and less aggressively percussive, and the overall 'feel' is more friendly and musical;
(b) I set a rather less marked cathedral reverb for all layers, so that the details are all just a little clearer, without losing worthwhile 'atmosphere'.

Unsatisfactory constraints on all the flies Nature-Symphonies are primarily that (a) the sequence of layer offset durations is fixed for the whole work, so every sound will repeat the same quasi-melodic / rhythmic pattern (no rhythmic variation apart from the flies' antics), giving a rather mechanical effect in many places, and (b) such a large number of layers means a quite high sum total of background noise, even after my best endeavours at noise-reduction — but at least from now on I've got a still more effective noise-reduction strategy.

I made the original recording in warm and quite humid weather, during a 4¼ hour session on 16 August 2024 on the top of Cranbrook Down (beside east limb of inner perimeter track of Cranbrook Castle, an ancient hill fort), high above the Teign Gorge, Drewsteignton, Devon, UK. This work uses the first half of that, the second half already being used for Nature-Symphony 72 (https://freesound.org/people/Philip_Goddard/sounds/766473/ ).

Advisory

Important! To get the best out of this, with its mass of detail, listen with high-grade headphones. This is particularly important here because, even with the greatly enhanced noise-reduction now, the high level of background sound (summed from all layers and amplified in the cathedral reverberation) makes the finer details more difficult to hear when listening with speakers (turning the volume up is NOT recommended because of some startlingly loud events (flesh-flies, Sarcophaga carnaria coming very close).

This recording taking place; furry windshield of recorder is the black thing among the sparse stunted bracken on left.

Techie stuff:

Recorder was a Sony PCM-D100, with two nested custom Windcut furry windshields. It was placed on a mini-tripod, set at a low height to reduce wind disturbance. Its mics were set at narrow angle (90°) for post-recording widening and thus zooming-in. In retrospect I'm doubtful that that was such a good idea, because that zooming-in would speed up a fly's passage across the soundstage, and thus shorten various notes, whereas I really want more longer notes.

Post-recording processing was to apply EQ in Audacity to correct for the muffling effect of the windshield, and of course the aforementioned widening, using A1 Stereo Control.

_________________________

Noise reduction

Noise-reduction procedure was in three stages, all in Audacity (the WavePad noise-reduction functions are all too prone to producing artifacts or removing too much, so I never use them):

• the recently added OpenVino AI noise suppression module, using the deepfilternet2 model and attenuation limit set at 7. Amazing that it could be so helpful (with care!) for my purpose, considering that it's modelled for extraction of voices / vocals, not natural soundscapes!
• Audacity's standard noise reduction function (spectral subtraction, using a submitted sample of 'clean' and constant unwanted background noise), set to 6dB noise reduction, with frequency smoothing bands set to 3 instead of my previously normal 6 (the default). This way I got the reduction without any noticeable artifacts at all — Wow!
• Because the two stages so far have reduced amplitude and made the sound over-bright, I correct for that (amplify +6dB, EQ curve a 9db linear tilt (8K to 100Hz away from the treble).
• Repeat the OpenVino noise-suppression, this time with attenuation limit set to 6, followed by another EQ tilt, this time just 3dB away from treble, and a little further amplitude increase.

The OpenVino processing increases dynamic range, so wouldn't always be a good option, but here that second pass in particular gets a really breathtaking performance from all my little winged friends.

I've found that in practice the OpenVino module, being AI driven, responds rather differently to each recording, so it looks as though for maximal non-destructive noise-reduction I'd have to experiment and find the right regime for each recording individually, which is really a time-consuming faff that I could do without!
_________________________

Layer pitch shifts (semitones above / below original): +5, +4, 0, -4, -15, -14, -11, -6, -5, -7, -11, -13;
Layer acoustic — all layers: nearer foreground in cathedral.

The pitch shift values given don't include the +3 semitones added to each in order to restore pitch after the slowing-down to slightly lengthen the notes.

Note that in this and all the other flies Nature-Symphonies I'm not changing musical speed of any layers; only pitch gets changed (using kHs Pitch Shifter Pro).

Please remember to give this recording a rating — Thank you!