necessary to manipulate the audio, a procedure carried out in the audio editing software “Reaper”, a Digital
Audio Workstation (DAW). The first manipulation technique used was gain increase, aiming to raise the
sound intensity. The gain implemented to the audio was +15 decibels (dB). However, due to the presence of
noise and hissing resulting from the material's degradation and state of conservation, the increase in gain
caused distortion in the audio (clipping). Thus, the first applied technique did not yet result in significant
improvements in the intelligibility of the voice and the piano.
To clean up the noise, it was necessary to apply a second technique, commonly known as “De-noise”, widely
used by the phonographic industry precisely for audio cleaning. For this purpose, the Spectral De-noise plug-
in, part of the “RX 9 software” package developed by “iZotope”, was used. The plug-in analyzes the spectral
profile of selected audio segments containing noise, processing the isolated segment, and allowing for its
cleaning without affecting other sounds or frequencies. In this process, some parameters within the plug-in
itself must be directly controlled by the user, such as smoothing levels and the gating level, which controls
how much of the original audio signal (with the noise) will permeate the obtained result. The applied
parameters were: Quality: D (Best); FFT size: 50; Multi-resolution on; Algorithm: extreme; Artifact control:
7.4; Smoothing: 10; Knee: 1.5; Whitening: 5; Enhancement: 5; Masking: 10; Threshold: 0; and Reduction: 12.
After the two audio treatment techniques, the enhanced phonogram was then ready for aural transcription,
which began with the transcription of the poem's text, now intelligible. In a second step, through internal
solfeggio, the rhythmic and intervallic relationships of the piece were noted down by hand, occasionally
using the piano to confirm the chords. Subsequently, the handwritten notes were transcribed into a score
editing software, generating a first version of the transcription.
From the beginning of the phonogram transcription process, the instrumental parts for the orchestral
version were used for verification purposes. This verification was especially important to complete the
transcription of the piano part (chamber version), which features very low and sustained notes, as well as
chords composed of very close notes, elements that hinder the precise definition of each note.
The first version of the transcription underwent a detailed revision, revealing some differences between the
phonogram and the transcription, especially in the piano part. The differences found were always subtle and
difficult to treat with precision. For this reason, a specialized technological resource was used to separate
the voice and piano audios into two distinct tracks, in the hope that this could aid the perception of the
details of each part individually.
This procedure was carried out using the software “StemRoller” (version 2.0.3), which can isolate vocal parts
from instrumental parts. As only a piano and a voice were involved, the procedure was simple: the track was
loaded into the software without any additional configuration being necessary. Two audio tracks were
created: one containing only the voice, and the other containing only the piano. The result was quite
satisfactory, as it was possible to hear the details of each part more clearly from these tracks.
The next step was to compare all the available materials: the enhanced phonogram with piano and voice
together, the first version of the transcription, and the piano and voice separated tracks. This procedure
boosted a second version of the transcription.
The first adjustments proposed to this second version are in measures 3 and 10, where it was possible to
identify, especially in the separated piano track, a high register musical line played on the instrument, which