Content uploaded by Rui Dias
Author content
All content in this area was uploaded by Rui Dias on Mar 06, 2014
Content may be subject to copyright.
A CONTOUR-BASED JAZZ WALKING BASS GENERATOR
Rui Dias
University of Porto
INESC-Porto
School of Applied Arts,
Polytechnic Instituteof Castelo Branco
(ESART-IPCB)
ruidias@ipcb.pt
Carlos Guedes
School of Music and Performing Arts,
Polytechnic of Porto (ESMAE-IPP)
INESC-Porto
carlosguedes@esmae-ipp.pt
ABSTRACT
This paper describes a contour-based algorithm for the real-
time automatic generation of jazz walking bass lines, fol-
lowing a given harmonic progression. A brief description
of the walking bass procedure will be presented, and also a
brief survey on some common implementations and tech-
niques.
This algorithm was implemented in the Max/MSP graphi-
cal programming environment.
1. INTRODUCTION
1.1 The Walking Bass Practice
The walking bass is a very common playing procedure
used in jazz music, in which the bass walks through the
chord and scale notes in a regular pulse. This is espe-
cially used when accompanying a solo, firmly setting the
base pulse like a metronome, and simultaneously exposing
the underlying harmony of the song. This procedure has
its roots deep in the first decades of the twentieth century,
and was developed through the years with bass players like
Jimmy Blanton, Ray Brown, Ron Carter and Charlie Min-
gus, amongst many others. A very good insight on the
evolution of the bass role and many of the key innovators
in jazz history can be found in The Jazz Bass Book, Tech-
nique and Tradition, by John Goldsby [1].
The basic idea of a walking bass line is to go from one
chord to the next, linking them by filling the middle beats
with notes of the chord or scale, typically describing a
smooth melodic line or pattern. Far from being completely
passive, however, the roll and behavior of the walking bass
can dramatically change from almost neutral smooth lines
to very abrupt register changes and energized rhythms, dy-
namically contributing to the overall group energy and mu-
sical result. The continuous flow of the regular walking
bass notes, together with its harmonic and melodic content,
form one of the most charismatic and important elements
in traditional jazz playing.
Copyright: c
2013 Rui Dias et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution 3.0 Unported License, which per-
mits unrestricted use, distribution, and reproduction in any medium, provided the
original author and source are credited.
The learning of this technique usually implies the learning
of melodic licks, for each of the commonly used chord pro-
gressions. These licks are small melodic phrases that are
particularly efficient and musical, and are usually related
both to the available notes of the current chord and un-
derlying scale and to physical placement of the notes and
fingers on the bass. Bass learning methods like the ones
by Ron Carter [2], Ray Brown [3], Bob Magnusson [4]
and Steven Mooney [5] lead the student through a series
of these licks, covering most common harmonic progres-
sions, in order to provide the student with the basic formu-
lae that can then be adapted to any song. As the walking
bass player develops his technique and gains more experi-
ence, however, he will be able to use these phrases more
articulately. More importantly, he will be able to move
away from them, intuitively creating more natural sound-
ing lines and integrating several other elements like orna-
ments, theme-specific elements and group feedback.
”(...) musicians string together a sequence of motifslicks
as they used to be calledmodified to meet the constraints
of the chord sequence. (...) Yet, the motif theory cannot be
the whole story.” [6]
1.2 Implementations
Due to the non-repetitive nature of the walking bass tech-
nique, its use in computer software is actually quite lim-
ited. Commonly, walking bass lines in use are pre-recorded
or manually written for the entire song length, whether
as audio recordings or MIDI events. This practice has
its roots in play-along recordings, like the widely known
Jamie Aebersold [7] long list of score + CD Play-A-Long
albums, with comping tracks recorded by real jazz mu-
sicians, allowing the practitioner to play-along with the
recording. Computer software facilitates this method by
easily allowing the independent mixing for each track, as
well as change the tempo and transpose an audio or MIDI
track, even on iOS devices with apps like Smudge Apps
Band [8] with multi-track recordings and mixer.
More advanced software use pre-recorded small phrases
for each chord-type and/or chord progressions, which are
then transposed and chained together according to some
more or less intelligent algorithm. This seems to be the
case with software like the extensive Band-in-a-box [9],
and more recently iReal b [10], on the iPad. This kind of
implementation can use audio or MIDI clips. While audio
Proceedings of the Sound and Music Computing Conference 2013, SMC 2013, Stockholm, Sweden
305
clips keep all of the little nuances, sound and groove of the
original player, the MIDI clips allow more flexibility for
editing notes, instrument, and even tempo and phrase ele-
ments.
These implementations based on the use of pre-recorded
phrases, whether audio or MIDI, have however some limi-
tations:
- If the number of pre-programmed phrases is small, the
output will easily sound repetitive;
- The larger the number of pre-programmed phrases, the
larger the chances of melodic inconsistencies and non-musical
results;
- In order to obtain smooth transitions between chords, the
pre-programmed phrases have to be very neutral, resulting
in a very neutral sounding bass line;
- It is not easy to handle less conventional harmonic pro-
gressions.
2. A CONTOUR-BASED APPROACH:
TECHNICAL DESCRIPTION
The work presented in this paper stems directly from the
research that led to the development of the GimmeDaBlues
app [11]. It describes the algorithmic generation of melodic
phrases that connect the chords in a previously defined har-
monic grid, by calculating a path from the current chord to
the next, according to user-defined settings controlling the
direction and range of the melodic contour.
The phrase generation algorithm consists basically in three
stages: Target Note calculation, Trajectory calculation, and
an event manager (Player). The general structure is showed
in Fig.1.
Figure 1. Algorithm structure.
2.1 Target note calculation
The algorithm needs to know the current chord and the
next one in order to be able to calculate a phrase. The
target note is the last note of the phrase to be generated. A
simple probabilistic algorithm chooses which of the notes
belonging to the next chord will be used e.g. fundamen-
tal, 3rd, 5th, 7th, etc.. Currently, in order to maintain a
strong sense of the base harmony, a setting of 100% prob-
abilities of choosing the fundamental note of the chord is
used. Then, according to the current note and to a direction
parameter, the algorithm will find the chosen chord note in
the right octave.
The direction parameter defines whether the target note
will be selected up or down, relatively to the current note,
and there are five different settings: lowest, down, nearest,
up, and highest. The down and up settings tell the algo-
rithm to search for the nearest note in that direction, while
with the lowest and highest settings, the algorithm will se-
lect the lowest and highest note in the instruments range.
This parameter can be defined manually or automatically.
So, for example, considering a double bass instrument de-
fined with a range from E0 to G3 (having C3 as the middle
C), if the current note is a C2, and the target note is an
F, the direction parameter will define which F will be se-
lected. The down setting would select F1, while the up
setting would select F2. The lowest note setting will return
F0 - the lowest F on the defined range - and the highest
setting will return F3 the highest F on the defined range.
Figure 2. Possible target notes for the four direction set-
tings.
With the ”nearest” setting selected, the algorithm will au-
tomatically go up or down, choosing the F that is nearest
to the current C2, which will be F2, because its a Perfect
4th interval, while F1 would be a Perfect 5th.
2.2 Trajectory
The trajectory is constituted by a selection of notes that de-
fine the path the bass line will take from the starting (cur-
rent) note to the final (target) note. In a typical case for
a chord duration of one bar in a 4/4 measure with the bass
playing quarter notes, the complete generated bass line will
have five notes, in which the fifth is the first note of the next
Proceedings of the Sound and Music Computing Conference 2013, SMC 2013, Stockholm, Sweden
306
measure. The three passing notes between the starting note
to the target note are calculated recursively, going from the
strongest beats of the measure to the weakest, and depend-
ing on the trajectory calculation algorithm.
An openness parameter will set how direct or indirect the
path will be, influencing the selection of the middle notes
in the calculated bass line. The name relates to the notion
of closed and open forms of a chord. The closed position
will be the more direct path to the target note, while more
open path will tend to use a wider trajectory, using chord
notes in an open form.
The notes to use will be drawn from chord tones, scale
degrees and chromatic inflections, according to each steps
beat position. The stronger beats of the bar will tend to
have chord tones, while the weak beats will tend to have
scale notes acting as passing notes from one chord note
to the next. The last beat of the bar can also be a chro-
matic approximation to the target note. This is a very com-
monly used technique, as the chromatic passing note cre-
ates a strong attraction to the target note, emphasizing it as
well as the sense of direction in the melodic phrase.
Fig. 3 shows four examples of possible trajectories of the
calculation of a bass line for a C7 F7 progression, where
the initial note is C2, and the target note is F2.
Figure 3. Phrase b) is the same as a) but with an ornamen-
tal triplet repeating the G in the third beat.
2.3 Ornaments
Although the construction of the phrases are the base of
the walking bass technique, there are several other aspects
regarding the notes, rhythm and articulations that have an
important role in a good performance. These aspects, here
referred to as ornaments, are little nuances and additions to
the phrases that dont change nor define the main contents
of the phrases, but nevertheless can contribute consider-
ably to the quality and the dynamic of the walking bass
lines.
The current implementation allows for the use of eighth-
note triplet variations (or eighth-notes with a swing feel-
ing) that can be set probabilistically. This is one of the most
common rhythmic variations in the walking bass technique,
in which some of the notes are anticipated by one triplet
(or swinged eighth) with the same or another pitch. The
control is done by a single percentage value, setting the
probability factor.
Figure 4. Phrase b) is the same as a) but with an ornamen-
tal triplet repeating the G in the third beat.
2.4 Control
The combination of the direction selection and the trajec-
tory openness provide a contour-based definition of the
walking bass line, which not only creates smooth and nat-
ural lines but also allows an effective and intuitive control
in interactive real-time implementations.
A simple example of a possible manual controller for this
walking bass generator would be a joystick type controller,
where the vertical axis would control the direction param-
eter, while the horizontal would control the openness pa-
rameter.
Figure 5. Example phrases with different settings.
Figure 5 shows three example phrases with different set-
tings:
- a) direction: nearest / openness: low;
- b) direction: down / openness: low;
- c) direction: down / openness: high.
3. CONCLUSIONS AND FUTURE
DEVELOPMENTS
The overall outcome of the described algorithm is quite ef-
fective and promising, mainly due to its flexible nature, by
adapting to any arbitrary harmonic progression and by al-
lowing a meta-control over the bassist behaviour.
Future implementations will focus on the way the phrase
generation can be controlled algorithmically, exploring the
contour shapes for the creation of motivic oriented groups
of phrases. Several strategies can be approached, like Markov
models and genetic algorithms, to control the succession of
phrase parameters musically.
Due to its recursive nature, the current algorithm only deals
with multiples 2/4, 2/2 and 4/4 measures. In metrical terms,
the beat hierarchy in measures with an odd number of beats
is not so clear, and thus the algorithm has to know how
Proceedings of the Sound and Music Computing Conference 2013, SMC 2013, Stockholm, Sweden
307
to handle them appropriately. Also, and relating to metri-
cal aspects in the phrase calculation, is the notion that the
phrases played by good bass players tend to have very of-
ten a length of two measures instead of just one. This cre-
ates more fluent lines contributing to the smoothness of the
musical form but also to the musical dynamics and move-
ment by relieving the measure-by-measure step size.
Regarding harmony, some interesting developments can
include the introduction of harmonic variations like chord
substitutions and alternate chord progressions, which is a
very common practice with advanced players, described in
books like the ones by Nettles and Graf [12], or Felts [13],
and addressed in the work by Steedman [14] [15].
Also, in order to make it sound more human and active, the
ornaments features regarding rhythm, dynamics and artic-
ulation will have to be addressed.
4. REFERENCES
[1] John Goldsby, ”The Jazz Bass Book Technique and
Tradition”. Backbeat Books, 2002.
[2] Ron Carter, ”Building Jazz Bass Lines”. Hal Leonard,
1998.
[3] Ray Brown, Ray Brown’s Bass Method”. Hal Leonard,
1999.
[4] Bob Magnusson, ”The Art of the Walking Bass”. Hal
Leonard, 1999.
[5] Steven Mooney, ”Constructing Walking Jazz Bass
Lines” (book series). Waterfall Publishing House,
2011.
[6] P. Johnson-Laird, ”How jazz musicians improvise”, in
Music Perception, Spring 2002, Vol. 19, No. 3.
[7] a list of the Jamie Aebersold
publications can be found at
en.wikipedia.org/wiki/List of songs in Aebersold’s Play-
A-Long series
(consulted in March 17th 2013).
[8] Smudge Apps - Band. www.smudgeapps.com
(consulted March 14th 2013)
[9] PG Music - Band-in-a-box.www.pgmusic.com
(consulted in March 17th 2013)
[10] Technimo LLC, iReal b. www.irealb.com
(consulted March 14th 2013)
[11] R. Dias, T. Marques, G. Sioros and C. Guedes,
”GimmeDaBlues: an intelligent Jazz/Blues player and
comping generator for iOS devices”. in Proc. Conf.
Computer Music and Music Retrieval (CMMR 2012).
London 2012.
[12] Barrie Nettles and R. Graf, ”The Chord Scale The-
ory and Jazz Harmony”. Rottenburg: Advance Music,
2002.
[13] Randy Felts, ”Reharmonization Techniques”. Boston:
Berklee Press, 2002.
[14] Mark Steedman, ”A Generative Grammar for Jazz
Chord Sequences”, in Music Perception 2, 1984. pp.
52-77.
[15] Mark Steedman, ”The Blues and the Abstract Truth:
Music and Mental Models”, in A. Garnham & J.
Oakhill (eds.), Mental Models In Cognitive Science.
Mahwah, NJ: Erlbaum 1996, pp. 305-318.
Proceedings of the Sound and Music Computing Conference 2013, SMC 2013, Stockholm, Sweden
308
