Comparative visualization of orchestral renditions of 'Esta Noche de Luna' tango composition

Digital information contained in audio files can be translated into graphical form as means to visualize the underlying structure of musical composition and as methodology to create abstract art. Visualization approaches include the use of variation in music’s loudness and frequency as input to imagery creation. Although music visualization is an active field of experimentation in the area of electronic music, its application to tango has been rare. Recently, Martin Calvino used sound mapping techniques and applied them to the visualization of 32 different orchestral renditions of the same tango composition (‘La Cumparsita’), performing a comparative visualization analysis based on loudness and frequency data.

In this study, the authors extended sound mapping approaches to the visualization of four different orchestral renditions of ‘Esta Noche de Luna’ tango composition. Differences in musical loudness among the renditions allowed us to visually differentiate their internal structure and variability. Variation in loudness levels among the renditions was also used as input for the creation of abstract artworks.

About ‘Esta Noche de Luna’ –

Esta Noche de Luna is a song dear to the heart of musicians, dancers and tango enthusiasts because of its melodic nature and romantic lyrics. The music was composed by Jose Garcia and Graciano Gomez, with lyrics from Hector Marco. It was first recorded in August of 1943 featuring singer Alfredo Rojas, and in December of the same year Carlos Di Sarli and his orchestra featuring singer Roberto Rufino, recorded a second rendition of the composition.

Jose Garcia directed the orchestra ‘Los Zorros Grises’ characterized by a rhythmic style; and

Graciano Gomez was a composer and orchestra director as well. Gomez reached prominence during the year 1955 when featured Uruguayan singer Nina Miranda as part of his ensemble. Hector Marco was a prolific writer and tango composer, having written many lyrics of tango songs.

In this work, we focused our attention to the following four renditions of ‘Esta Noche de Luna’:

1943 Graciano Gomez & Jose Garcia featuring singer Alfredo Rojas

1943 Carlos Di Sarli featuring singer Roberto Rufino

1944 Francisco Canaro featuring singer Carlos Roldan

1955 Osvaldo Pugliese featuring singer Jorge Maciel

The lyrics of the song is structured in four paragraphs with some orchestras leaving out the second and four paragraphs all together.

Acercate a mi y oirás mi corazón contento latir como un brujo reloj. La noche es azul, convida a soñar, ya el cielo ha encendido su faro mejor. Si un beso te doy, pecado no ha de ser; culpable es la noche que incita a querer. Me tienta el amor, acércate ya, que el credo de un sueño nos revivirá. Corre, corre barcarola, por mi río de ilusión. Que en el canto de las olas surgirá mi confesión. Soy una estrella en el mar que hoy detiene su andar para hundirse en tus ojos. Y en el embrujo de tus labios muy rojos, por llegar a tu alma mi destino daré. Soy una estrella en el mar que hoy se pierde al azar sin amor ni fortuna. Y en los abismos de esta noche de luna, sólo quiero vivir, de rodilla a tus pies, para amarte y morir. Corre, corre barcarola, que la luna se escondió.

In the case of Canaro and Pugliese, their interpretations repeat the third paragraph but in Canaro’s version only the last part of the third paragraph is repeated. Because the voice of the singers can be considered an additional instrument, variations in the repetition of the lyrics contribute to loudness differences among the renditions and thus affect the visual pattern as well.

Visualizing the underlying architecture of ‘Esta Noche de Luna' based on loudness as data input

Changes in music loudness were used as input to create graphics that could provide an integrated visualization of the composition and its comparison among the four orchestral renditions analyzed in this study. For this an algorithm was created that mapped loudness level throughout the entire composition to the length of vertical lines, the taller the line the louder the music at that point in time. When loudness levels exceeded 15%, vertical lines were painted orange relative to grey lines for all loudness levels inferior to 15%. In order to appreciate differences in loudness levels that were superior to this threshold throughout different moments of the composition, loudness was mapped to variations in the stroke weight of orange vertical lines, wider lines depicting higher loudness levels. This means that for loudness higher than 15% two mapping approaches were applied, the length of the vertical line as well as its width. Results are shown in Figure 1 and Figure 2.

Figure 1. Sound mapping of four renditions of ‘Esta Noche de Luna’ tango composition. Loudness levels throughout the composition are represented by vertical lines, with the length of each line mapped to loudness (louder sound equals longer lines). The graph reads as written text, starting at the top-left (beginning of the song) and reading from left to right and from top to bottom, with the end of the song at the bottom-right section of the graph. Loudness levels superior to 15% are shown as orange lines with stroke weight mapped to loudness levels as well (wider lines depicting louder levels). Loudness levels below 15% are displayed as vertical lines of grey color. Visualization of four orchestral renditions are shown: Graciano Gomez and Jose Garcia, Carlos Di Sarli, Francisco Canaro, and Osvaldo Pugliese respectively. Click on figure to interact.

Figure 2. Graphs previously shown on Figure 1 were filtered for loudness levels below 15% and thus were not drawn. The figures shown here only display loudness levels above 15% as orange lines. Different patterns of loudness are clearly appreciated among the four renditions. Click on figure to interact.

This analysis provided us with an overall understanding of the differences among renditions, with Garcia and Di Sarli’s distributed loudness levels across the composition, whereas Canaro’s renditions displaying very limited variation in loudness level. The rendition of Pugliese was perhaps the most interesting because it displayed punctuated outbursts of loudness as well as periods of very low loudness levels.

Punctuated outbursts of loudness levels in Pugliese’s rendition compared to the other three renditions can be clearer addressed using a different visualization approach in which the four renditions are visualized at once by mapping dots drawn based on loudness levels (Figure 3 and Video 1).

Figure 3. Comparative analysis of four renditions of 'Esta Noche de Luna' are shown on the same graph as they were played simultaneously (see Video 1). Songs start at the left end of the canvas and transition towards the right end. Loudness levels at any given time are represented as colored dots: Garcia-Gomez (light orange), Di Sarli (light pink), Canaro (light violet), and Pugliese (light blue). The positions of dots along the y-axis are dependent on loudness, with lower dots representing louder sounds. When loudness levels for each song exceeded 15%, dots were painted white.

Video 1. Dynamic visualization of Figure 3 is shown, with four renditions played simultaneously.

From top to bottom: Garcia-Gomez, Di Sarli, Canaro, and Pugliese renditions, respectively.

Variation in loudness levels among renditions as data input for the creation of computational abstract art

It was interesting for us to explore variations in loudness among different renditions of the same tango composition as data input for the creation of visual novelty. Three ideas for abstract visualization were explored and the results for each are presented. The first approach was based on exploring visuality derived from images shown on Figure 2. We considered they provided a sufficiently appealing esthetic value and thus colored lines from each composition with the same color pattern as Figure 3, presenting loudness levels above 15% from all compositions within the same artwork (Figure 4).

Figure 4. Artwork created using variation in loudness levels above 15% among the four renditions analyzed in this study. Color code is the same as in Figure 3.

A second approach that integrated variation in loudness coupled with randomness was used to create a piece that explored concepts of visual alignment among renditions, drawing lines connecting rendition's paths when differences in loudness levels among them exceeded 10%. The integration of a random component in the creation of visuality assured that each time the algorithm was run, it produced a different outcome that potentiated loudness differences. Each run was then placed under different blue backgrounds as shown in Figure 5.

Figure 5. In this artwork five different runs of the same algorithm are presented under different blue backgrounds. Renditions are shown as colored paths starting at the center-top and running towards the center-bottom of the canvas. When differences in loudness levels among renditions exceeded 10%, colored lines connected the path among pairs of renditions (Garcia-Gomez & Di Sarli, Di Sarli & Canaro, Canaro & Pugliese, respectively).

The third and last approach explored also implemented an integration of loudness levels for each rendition and a randomness component. Here, the combination of loudness levels and randomness affected the direction of the path along the x-axis as well as the y-axis. The artwork included the four renditions as they were played simultaneously (just as in Figure 5) with each run of the algorithm depicting a different starting point for each song. Here, colored lines were also drawn together with colored circles each time loudness among pairs of renditions exceeded 10%. The color and transparency of the stroke for each rendition's path was determined based on the position within the canvas. Visual results from four runs of the same algorithm are presented (Figure 6), with each run having a different starting position for each rendition. Color code is the same as in Figure 5.

Figure 6. Four abstract artworks responsive to loudness levels from four renditions of 'Esta Noche de Luna' are shown. Colored lines and circles represent loudness differences above 10% among pairs of renditions: Garcia-Gomez & Di Sarli, Di Sarli & Canaro, Canaro & Pugliese, respectively. Click on figure to interact.

Art created from music has emotional value

It is interesting to note that although the artworks presented here can be considered abstract, they are in somehow representational as well because they still retain a connection to data input from music files. They also have an added emotional value as they evoke musical compositions that we care about, not only in terms of the music per se but also from a social and cultural perspective. For instance, the authors danced together one of the renditions presented in this study, and from there arose the idea of this collaboration. Knowing the artworks represent in part renditions of a famous tango composition also brings in the viewer memories of his home country (Uruguay) and the past, connecting the artwork to the artist in the most personal and deepest manner.

About the authors and their contributions

Meg Farrell is an Adjunct Professor at Medgar Evers College, CUNY. Meg's contributions to this work included the original idea of visualizing 'Esta Noche de Luna' tango composition. Meg Farrell contributed with the Pugliese's rendition of the song.

Martin Calvino is a new media artist and creative technologist interested in the intersection of art, technology, tango and science. Starting this Fall of 2017, he will join the Interactive Telecommunications Program (ITP) at Tisch School of the Arts, New York University. He also holds a PhD and MS in plant molecular genetics. Martin's contributions to this work included all the visualizations of 'Esta Noche de Luna' tango renditions, all the artworks presented and the first written draft of this text. Martin's contributed with audio files for the first three renditions of 'Esta Noche de Luna'.

References –

Hector Marco’s autobiography: (accessed July 23, 2017). (accessed July 24, 2017). (accessed July 24, 2017). (accessed July 24, 2017).

Keywords: Martin Calvino; abstract art; computational abstract art; sound mapping; sound visualization; algorithmic art; comparative visualization; tango music; tango.