Cochoreo: A Generative Feature in idanceForms for Creating Novel KeyframeAnimation for Choreography

Kristin Carlson, Philippe Pasquier, Herbert H. Tsang, Jordon Phillips, Thecla Schiphorst and Tom CalvertSchool of Interactive Arts and Technology, Simon Fraser University, Canada

Applied Research Lab, Trinity Western University, Canadakca59, pasquier, htsang, jjp1, thecla, tom@sfu.ca

Abstract

Choreography is an embodied and complex creativeprocess that often relies on ‘co-imagining’ as a strat-egy in generating new movement ideas. Technologyhas historically been used as a tool to augment cre-ative opportunities in choreographic process, with mul-tiple choreographic support tools designed to functionas a ‘blank slate’ for choreography. However, few ofthese tools support creative authoring with interactiveor generative components. Cochoreo is a sub-modulefor generating body positions as keyframes that catalyzecreative movement, as part of the movement sketchingtool idanceForms (idF). Cochoreo catalyzes movementsketching by using parameters from Laban MovementAnalysis, an existing movement framework, to gener-ate unique keyframes that are used as seed material forchoreographic process. idF is a creativity support toolthat engages with choreographers’ creative movementprocess by design. This paper presents the design ofCochoreo and evaluations from our pilot study with uni-versity dance students.

IntroductionChoreographers are artists who are always searching for newinspirations from which to design novel movement ideas.They derive inspiration by exploring movement physicallyon themselves, they view movement on others, they observeinteractions between strangers, explore the physics of inan-imate objects and manipulate existing technology to createnew movement experiences. It is in these exploratory inter-actions that choreographers not only discover ideas but iter-ate them to develop larger pieces of creative movement ma-terial. The performance theorist Andre Lepecki developedthe term ‘co-imagining’ for these specific kinds of interac-tions that require multiple participants to devise and developideas, but who are not necessarily co-authors in the composi-tion process (Cunteanu 2016). This paper discusses the cur-rent state of choreographic support tools and how our sys-tem, titled Cochoreo, addresses existing gaps between thedomains of creativity support tools and autonomously cre-ative systems.

While there are a variety of digital systems designed toengage with choreographic process, few support inspirationof new movement ideas or the iterative process of develop-ing movement material. Current tools fall on a spectrum of

possible choreographer interaction, with limited options forco-imagining systems. Creativity support tools aid a chore-ographer in their existing creative practice, yet do not offernew movement ideas to the choreographer. Autonomouslycreative systems generate novel movement options but donot have a way to iteratively interact with a live choreogra-pher. Few co-imagining systems exist yet none support thechoreographers personal exploration of novel movement.

To combine the functionality of a creativity support toolwith an autonomously creative system we have designed Co-choreo to co-imagine novel movement with the choreogra-pher. Cochoreo is a sub-module within the existing plat-form idanceForms, a sketching tool for movement basedon creating and animating keyframes. Keyframes are sin-gle frames, taken from film terminology and used in ani-mation to describe important start and stop points (see Fig-ure 1). Cochoreo generates keyframes (as single frames ofbody positions) and interpolates between keyframes to ani-mate choreographer-designed movement.

Cochoreo generates novel keyframes for body positionsby using a fitness function designed and tested by a chore-ographer. The iterative design process by the choreographerensured that generated body positions would be

Figure 1: Keyframe Layout in idanceForms

385

380Proceedings of the Seventh International Conference on Computational Creativity, June 2016

unfamiliar, unstable and utilize complex movement un-derstanding. There is also a parameterized fitness functionoption so that the choreographer can adjust generation op-tions based on their personal preferences. The Cochoreokeyframes can then be edited and manipulated manuallywithin the idanceForms framework.

Cochoreo was designed to leverage a co-imaginative ap-proach to embodied choreographic process in technology.Cochoreo reflects the use of chance procedures made fa-mous by world-renowned choreographer Merce Cunning-ham, who used the historical system DanceForms in hischoreographic process. (Schiphorst et al. 1990). Our teamhas re-designed DanceForms (to idanceForms) to functionon a mobile platform which utilizes affordances of a tabletfor capturing and manipulating movement data.

This paper discusses the gap between creativity supporttools and autonomously creative systems and illustrates anaddressable gap in the design of choreographic supporttools. We describe the system design of Cochoreo andpresent a pilot study with novice choreographers that ex-plores their experience of choreography in relation to theintegration of idanceForms, a platform for sketching move-ment and its generative feature Cochoreo.

BackgroundThere are a variety of computational creativity projects thatexplore the generation or augmentation of movement ma-terial for choreography. Few of these systems are creativeon their own and most involve some level of interactionwith the human creator. However, these interactive sys-tems often do not provoke creative compositional choicesin the creator, and do not support the “sketching process”.For example, programs such as Adobe Photoshop or Mi-crosoft Word give artists a “blank slate to put their ideas onbut do not assist them artistically in their practice (Cough-lan and Johnson 2009). We are interested in how an au-tonomous creativity component can support the creative pro-cess of the choreographer, to enable co-imaginative inter-action. In order to implement techniques that engage theagency of choreographers, we illustrate a selection of exist-ing systems that support choreographic process (see Figure2). A deeper analysis of prior work includes the survey pa-per by (Fdili Alaoui, Carlson, and Schiphorst 2014) thatdescribed existing systems which have been developed todigitally reflect on movement material, to generate choreo-graphic material, to provide real-time interaction with move-ment material, and to annotate movement material.

Systems that interactively support generative techniquesand choreographic process include The Dancing Genome

Project, Web3D Composer, Viewpoints AI and the stan-dalone idanceForms. The Dancing Genome Project devel-oped a genetic programming model to explore sequences ofmovement in performance (Lapointe and poque 2005) (La-pointe 2005). The system analyses movement data and re-organizes it to create a new sequence with the same move-ments. This system was used to generate variations of move-ment phrases which were performed by a combination oflive and digital performers. Web3D Composer creates se-quences of ballet movements based on a predefined library

Figure 2: Co-Imagining Systems Scale

of movement material (Soga et al. 2006). The system allowsthe user to select movements from a pool of possibilities,which shift based on structural ballet syntax. This system isused mainly as a teaching tool to support the developmentof ballet structure knowledge. The Viewpoints AI projectused the Viewpoints compositional framework to create areal-time interactive system exploring dance improvisationstrategies (Jacob, M, Zook, A, and Magerko, B 2013). Thesystem used kinect data and the SOAR reasoning frame-work to create a repository of short and long-term memoryof the choreographers movements that select and apply dif-ferent response modes and improvisational strategies. Boththe system and the performer attend to each other’s move-ment choices by interactively improvising movement mate-rial. idanceForms enables choreographers to design move-ment poses as keyframes and then animates them, creat-ing an iterative and reflective space for choreography design(Carlson et al. 2015a).

Systems that border on autonomous creativity includethe Cochoreo, Tour, Jete, Pirouette, DANCING and StyleMachine systems. Cochoreo generates body positions askeyframes within the idanceForms sketching application,to be used as catalysts for innovative movement design.Keyframes are animated and can be edited and sequencedwithin the idanceForms platform. Making use of the Dance-Forms framework, Yu and Johnsons system generates au-tonomous movement sequences through the use of a Swarmtechnique in their project titled Tour, Jete, Pirouette (Yuand Johnson 2003). This project used the existing li-braries of movement within the DanceForms software toautonomously generate sequences from a series of individ-ual movements onto a group of dance avatars. This cre-ated group movement sequences explored by choreogra-phers who deemed the movement too challenging to performexactly as the system did. DANCING used a series of music-related parameters, spatial pathway rules and a predefinedlibrary of traditional movements to generate Waltz choreog-raphy using a Genetic Algorithm (Nakazawa and Paezold-Ruehl 2009). By connecting the correct, predefined ‘steps’in a domain-specific sequence that provides stage directionsand orientations, this system generates syntactically correctmovements in a complete choreography that are representedas ASCII (American Standard Code for Information Inter-change) symbols on a birds eye view of the stage. It wasnoted that the generated choreography was able to be per-formed by ballroom dancers. Brand and Hertzmann devel-oped a system called Style Machine that generates stylis-tic motion by using unsupervised learning techniques based

386

381Proceedings of the Seventh International Conference on Computational Creativity, June 2016

on a Stylistic Hidden Markov Model (SHMM) (Brand andHertzmann 2000). This model learns patterns from a highlyvaried set of movement sequences recorded from motioncapture data. The model then manipulates movement byidentifying structure, style and accidental properties and ap-plying style qualities to movement (such as modern dancestyle in ballet movements). Alemi, Li and Pasquier devel-oped an interactive agent model that can capture and con-trol the affective qualities of movement patterns (Alemi,Li, and Pasquier 2015). They trained a Factored, Condi-tional Restricted Boltzmann Machine (FCRBM) with a cor-pus of movement captured from two actors that was anno-tated based on their arousal and valence levels.

This selection of systems illustrates the developments to-wards interactively co-imagining choreographic material be-tween a system and a human, yet there continues to be a gap.

DanceForms HistoryUsing digital tools to support the creative process of chore-ography has a historical precedent. DanceForms (formerlyLife Forms) is a human figure animation system that is op-timized for dance (Calvert et al. 1991) including the samecapabilities that are available in general purpose animationsystems (e.g. Maya, MotionBuilder, 3D Studio Max, orUnity)(see Figure 3). While the system has been used bymany choreographers, the most well known is Merce Cun-ningham. Cunningham used DanceForms to design move-ment as inspiration for constructing dances, exploring therandom and procedural components of the system into hisexisting creative process using Chance Operations (a versionof controlled randomization for content selection). Cunning-ham is a seminal figure in the history of choreography world-wide, and a unique ’user’ of technology in dance, in partic-ular the DanceForms software.

The Life Forms / DanceForms software was designed foruse with desktop or laptop computers and normally requiresa large screen (Calvert et al. 1993). Typically the user in-teraction requires that up to 5 windows be open at any time.The computer and the screen can be used in a studio but thecomputer is typically seen to be cumbersome and does

Figure 3: DanceForms Interface

not merge easily into a mobile, in-situ movement practice.The great advantage of mobile devices is just that: they aremobile. They can be carried onto the dance floor and theanimated movements compare directly to the movements ofthe live dancers. There are also new affordances in mobiledevices that we can take advantage of; the use of accelerom-eters to determine the acceleration, velocity and position ofa limb and the use of an integral camera to capture the stanceof a live dancer.

DanceForms has three views: space, time, and body-position. The space view allows the user to design move-ment pathways as spatial patterns. The timeline allows thechoreographer to design sequences and timings of move-ment. The body-position view allows the user to designbody positions using joint manipulation or to choose cod-ified positions from pre-designed libraries. Libraries weredesigned by using the corpus of standard positions that de-fine a movement language in techniques such as ballet ormodern. These Danceforms libraries have been used assource material in generative composition using a Swarm al-gorithm to automatically compose sequences of movement(Yu and Johnson 2003). While DanceForms is the mostarticulate system available for computer-supported chore-ography, its precision-based design does not support rapid,portable, mobile, embodied or experiential forms of inter-action. However, the rich foundation DanceForms providesfor supporting movement design in software is highly use-ful as a step for mobile development and exploration ofmovement-based sensors for sketching choreography.

Cochoreo and the idanceForms PlatformCochoreo is a sub-module of the idanceForms (idF) plat-form, a tablet-based mobile animation tool. This section willdescribe the idanceForms platform first, and the Cochoreodetails second to illustrate the platform in which Cochoreooperates.

idF is a creativity support tool that allows the choreogra-pher to sketch movement by creating, editing and viewinghuman figure animation on a tablet (Carlson et al. 2015b).idF differs from DanceForms in many ways: idanceFormsis designed to support the sketching process of choreogra-phers and is not meant to support the highly detailed tradi-tional process in DanceForms. The shift in interaction frommouse-based to touch has dramatically changed the designto be more minimal but directed towards a choreographerworking in an embodied way. This inspired the develop-ment of the Camera Keyframing feature, where a snapshotof a live dancer can be taken and used as a keyframe. id-anceForms has been designed based on the epistemologyof choreography; leveraging whole-body interaction as wellas the playful and low-risk properties of sketching to cre-ate a mobile support tool for exploring creative movementin-situ (Blom 1982) (Studd and Cox 2013). By using an an-imation platform we can continue to provide an element ofprecision that the original DanceForms software maintainswhile opening to new opportunities for interaction, designand representation of movement. Our contribution with iDFis its application to the live, in-situ creation and iteration ofcreative movement.

387

382Proceedings of the Seventh International Conference on Computational Creativity, June 2016

Figure 4: idanceForms Viewer and Skeleton Editing Tool

The ‘home’ screen is a playback screen that enables thechoreographer to view the animation on a ‘stage that theycan move around using single finger touch to rotate aroundthe space as well as pinch gestures to zoom (see Figure4). The playback view is an important piece of the choreo-graphic process, because it provides opportunities for view-ing the animated movement, understanding the movementthrough the kinesthetically empathetic experience and re-flection on the selection of and sequencing of still forms askeyframes. Playback is the portion of the creative processthat provokes reflection and evaluation of choices made inthe sketching process. Playback is the result of rapid proto-typing: creating a space for choreographers to reflect in ac-tion and quickly continue working to create personal mean-ing.

Sequencing KeyframesOnce the choreographer has captured still poses to use askeyframes in their animation they have options for adjustingsequencing and timing of keyframes. Touching a keyframeonce will select it and enable dragging and dropping to re-order keyframes for designing creative sequences. Becausewe are working with keyframes, there is built-in linear in-terpolation that takes the shortest path to move from onekeyframe to the next. This creates a unique ‘movement fromthe transition between a starting and ending still pose. Thechoreographer can control the timing of this ‘movement byadjusting the timing into and out of a keyframe with the tim-ing bar at the top of the editing screen.

Skeleton Editing ToolUsing the finger gesture the user can manipulate the skeletonin a joint and limbs level (see Figure 4). This fine controlis facilitated by the gimbal ball visualization where the usercan select the axis of the movement and then move the limbsaccordingly.

Data RepresentationThe skeleton setup and skinning method we use is basedon the COLLADA standard, using the CMU motion captureskeleton (cmu ). We use a linked list of keyframes to

Figure 5: Camera Keyframing Feature

store our animations for two reasons a) this makes it eas-ier to swap or move keyframes around during editing and b)it is also faster to play back the animation. Each keyframestores a pose and an integer representing the number of in-between frames until the next keyframe. We do not store theexplicit frame number in a keyframe, this is determined bythe sum of the previous keyframes in the linked list addedto the sum of inbetween frames for each keyframe, and thisallows keyframes to be easily swapped / moved without re-calculating their frame.

Camera KeyframingidanceForms has developed a camera keyframing feature toenable embodied forms of interaction. Utilizing the 2D cam-era in mobile devices, the background is removed and thedancers still pose is compared to an existing database of im-ages with existing skeletal data (see Figure 5). The built-incomputer vision algorithm will then capture the pose andsearch through a database of pre-stored standard poses inorder to try to find a corresponding skeleton pose. Once theskeleton pose has been found, it will be added to the list ofkeyframes. We have designed a database of movement us-ing planar poses that can be easily detected from the frontwithout occlusion. These poses include general and creativebody positions as well as an imitation of alphabet letters thatwas used in a prior study with youth (Carlson et al., 2015).The existing skeletal data is used to create a keyframe thatcan be added to a sequence of keyframes to create an anima-tion of movement and be further manipulated by the chore-ographer. This ‘capture process is an exciting innovation formovement interaction which enables us to capture still formsas a wide range of potential planar figures.

Cochoreo System DesignCochoreo is a sub-module of idanceForms, used to gener-ate novel keyframes for creative movement (see Figure 7).Keyframes consist of a still body position, a human-shapedavatar with movement possibilities in all 3 axis. Cochoreouses a Genetic Algorithm to evolve new keyframes from an

388

383Proceedings of the Seventh International Conference on Computational Creativity, June 2016

initial gene pool. Cochoreo is an extension of the Scuddlesystem (Carlson et al 2011), a generative system for move-ment catalysts. Scuddle generated movement catalysts asstatic positions with performative instructions to be inter-preted by a choreographer and used as inspiration for de-signing new novel movement.

Cochoreo’s implementation in the idanceForms animationplatform enables a new parameterized fitness function, en-gaging the choreographer in the generative design process.idanceForms in return provides a sequencing and animationplatform to iteratively view and design phrases of movementwith the user, creating files that can be documented and usediteratively throughout the choreographic process. CurrentlyCochoreo operates using 2D data with the z-axis zeroed out.While we plan to move to 3D in the future, it will require an-other iterative design process to develop a constraint systemfor preferred creative catalysts.

Genetic AlgorithmWe use a Genetic Algorithm to evolve movement catalysts.This approach enabled us to control fundamental compo-nents that problematize the choreographers process of cre-ating movement, while generating novel inspirations formovement solutions. Genetic Algorithms are typically usedto explore a wider range of potential solutions than othersearch algorithms can (Russell and Norvig 2010). We gener-ate a population of 500 random individuals and give a scorefor their fitness against the prescribed goals for success. Thisinitial population is then subjected to an iterative cycle of se-lection and breeding. Genes are bred using a two point crossover function with a 10 percent mutation percentage to cre-ate a new population that maintains diversity. Once a cycleis complete the new population is judged on its fitness onceagain and the process continues for a fixed number of fiveiterations or until a certain fitness threshold is reached (Flo-reano 2008) (Russell and Norvig 2010). More details on thegenerative process can be found in the Scuddle system pa-per (Carlson, Schiphorst, and Pasquier 2011).

Fitness FunctionsCochoreo has two fitness function options to evaluate novelbody position criteria in keyframes. The options are: a pre-defined fitness function and a parametric fitness functionbased on Bartenieff Fundamentals movement constructs.The pre-defined fitness function uses a set of criteria specif-ically for provoking novel keyframes based on traditionaldance movement. We have developed heuristic rules basedon movement patterns discussed in Bartenieff Fundamentalsand the authors expertise in contemporary dance practiceto inhibit traditional habits when creating movement (Studdand Cox 2013). The fitness function evaluates each catalystcomponent separately (body symmetry, body position andlevels) and then calculates the overall score. Preferred posi-tions are those that highlight contralateral movement (bodyasymmetry), unstable levels with partially bent joints to cre-ate novel movement options. More on this function can befound in our prior paper (Carlson, Schiphorst, and Pasquier2011). The parameterized fitness function allows the

Figure 6: Pre-Defined Fitness Function Interface

choreographer to change the weighting of each parameter,creating more personalized generated options.

Cochoreo InteractionTo use the Cochoreo feature in idanceForms, the choreogra-pher goes into the keyframe editing screen and selects a newkeyframe. The Cochoreo screen is simple, providing a but-ton for ‘Generate and a gear icon for access to the settings(see Figure 7). Every time the Generate button is presseda new keyframe is created. This keyframe can then be re-edited in the skeleton editing view. Limbs can be isolatedby selecting them, changing the color from pink to white.Isolated limbs will stay in place during the next generationcycle and can be un-selected by touching them again. Ifgeneration including a spinal configuration or spatial orien-tation is desired the user can manipulate these features firstand then generate new keyframes in which the edits will beretained.

The default fitness function generates keyframes based onthe pre-defined rules, developed through an iterative designprocess to specifically restrict habits from dance techniqueand provoke novel movement options. This default fitnessfunction weights body asymmetry, uneven reach space andunstable levels more strongly to encourage novel movementexploration.

The parameterized fitness function enables the choreog-rapher to select options based in the Bartenieff Fundamentalparameters to weight the probability of that feature more orless strongly (see Figure 8) (Studd and Cox 2013). Parame-ters include: Body Half (symmetry on one side of the body),Upper Lower (symmetry on top or bottom half of the body),Cross Lateral (symmetry across the body with one arm andone leg), Near Reach Space (arms contracted), Far ReachSpace (arms extended), Knee Extension/ Flexion (creatingmore or less stable levels).

Body positions can also be interacted with in the gener-

389

384Proceedings of the Seventh International Conference on Computational Creativity, June 2016

Figure 7: Parameterized Fitness Function Interface

ation by isolating limbs (see Figure 9). By isolating limbsthey are removed from the algorithm while remaining limbscontinue to be generated. The shape of the spine can also bemanipulated by manually editing the vertebral joints usingthe skeleton editing features (selecting individual joints andmoving them using the 3 axis) and then generating new limbpositions.

Choreographic Study Exploring CreativeExperience

We evaluated the system in a pilot study with 14 novicechoreographers who were second year university dance ma-jor students. Choreographers met for two workshops over aweek and had a composition assignment in-between

Figure 8: Isolating Limbs in Generation

workshops. The technology was introduced as a tool tosupport their existing choreographic process, and they wereguided through short exercises to use it while constructing amovement phrase. Observational data was collected by theresearcher through notes, photo and video documentation.Semi-structured focus groups were used to gather informa-tion about the creative experience. Data was analyzed usingthematic analysis to highlight salient topics identified aboutchoreographer’s creative experience.

The goal of this research was to explore how choreog-raphers can interactively develop creative movement with asystem, where both user and system generate creative ideasand iteratively develop a movement phrase. Understandinghow choreographers would work with the system requiredobserving and understanding the embodied process of ex-ploring and ’trying on’ the movement on their particularbodies. When provided with the idanceForms app, choreog-raphers explored the shape of the movement on their body.They then made mapping decisions about how to translatethe data from the avatar to themselves, exploring it on thebody and finding new connections where they could inter-actively augment the movement design themselves. In it-eration with the system choreographers would insert newmovements, and augment existing movements by either re-capturing the movement into the device (and manipulatingit there) or by viewing the movement ‘cues from a differentperspective.

Novelty of Generated KeyframesCochoreo’s generated movement catalysts were viewed tobe interesting, suggesting movement options that the chore-ographer would not have developed themselves. Paired withthe manipulation tools of idanceForms, choreographers hada variety of options for controlling the generation of move-ment material. The camera keyframing feature enabledchoreographers to capture positions with the iPad’s camera,which was matching images to an existing database (and notalways precise to the movement performed by the choreog-rapher). However, when the data was less precise and moreembodied it supported the choreographer’s exploration ofmovement.

D: I liked working with the program because it pushedme to do movement I would never think of ... that wasreally interesting to me to try to put myself in this un-comfortable place and now a week later be comfortablein moving in that sort of way.

Resolution of Data/ Mapping StrategyIn the original design of Scuddle, generated positions werestatic and minimalistic stick figures. This design promptedchoreographers to focus more on the interpretation of figures(and invention of new positions) than attempting to map theexact 2D position onto their moving 3D bodies. The lowresolution of data catalyzed novel exploration, yet it couldnot guide positions into movement. Cochoreo generateskeyframes, yet they are animated to create movements anduse a 3D stick figure. This design uses a higher resolution ofdata, which prompts more attention to the physical mapping

390

385Proceedings of the Seventh International Conference on Computational Creativity, June 2016

Figure 9: Pilot Study Session 2

from screen to body. When choreographers were learningmaterial from the static keyframes, the higher resolution ofdata prompted them to focus less on their creative interpre-tation of the positions themselves, but brought attention tothe transitions between positions as they had to maneuverdynamic changes in the data. Though when choreographersattended to the animated phrases instead of the static posi-tions, they attended to the translation of dynamic parameterssuch as time and momentum more than the body position.

B: I felt that it felt better on my body if I used the appfor inspiration for my movement and didnt necessarilytry to replicate it exactly. So that really helped. Andthen when we sped up our movement or added repeti-tion that helped it flow more easily through my body.

B: I felt that after going back and forth, like when Iwas first working with it it was very planal, but oncewe took it from there and took the movement home Icould explore the other aspects of it. So then today inthe space, even though we went back and added that bitin, I tried to keep the idea of my home movement butfrom a generated source.

Co-Imagining/ Interaction with SystemChoreographers were asked to generate multiple keyframesin Cochoreo, then learn them on their own body to cre-ate a movement phrase. The goal was to eventually movesmoothly back and forth between designing movement inthe system and exploring the movement on the body. Whilethe translation of data from device to body and back wasa new challenge for many choreographers, they discoveredunique perceptions to movement design through their inter-action with the system. These included an attention to mo-mentum as related to time in the system, the focus on angularlimb positions and how they moved through time and atten-tion to spatial orientation and engagement in relation to a fo-cus on a mobile device. Choreographers also became awareof their movement habits (many which developed throughmovement training) and preferences when exploring move-

ment with specific sensory feedback (if a movement ’feelsright’ or ’looks right’ in the mirror).

A2: It was interesting working with this movementaway from an image (the software) because I feel likewhen you have this image in front of you, the mirrorneurons you want to mimic this movement, and thatsthe way we learn movement, rather than learning fromfeelings, so not having that ’mirror’ (of an image) andtaking what this movement was in our memories andplaying with the feeling of what it was, it was interest-ing and illicited new... it helped me evolve the move-ment. Having the exposure to it and then taking it away.

J: I usually focus more on momentum so its interestingto approach with more emphasis on the angles, becauseits like whoa I have limbs! I just realized I have limbsand its in my face! Also realizing peripheral vision be-cause there is a lot of stuff with angles happening backhere which I dont usually think about.’

ConclusionThe evolution of Cochoreo as a sub-module within the id-anceForms framework enabled us to explore how a creativ-ity support tool could also provoke creative choreographicchoices. We observed how choreographers devise move-ment using embodied methods, and augmented that processby inserting Cochoreo phrases within the embodied meth-ods.

We view Cochoreo as a preliminary exploration of gen-erative authoring tools for movement to evaluate how theaffordances of such a system can support the creative valuesof a choreographer. While the goal of this project is to createan interactive toolkit for choreography design where a work-flow can move smoothly between the choreographer and thetechnology, this is a complicated process that does not haveobvious solutions in the near future. We are interested inhow to take small steps to work towards this goal. In thisstudy we observed the playful discovery process that eachchoreographer experienced and began weaving into craftedmovement sequences. We see potential for systems that uti-lize generative movement augmentation to create embodiedand personalized qualities of work, as opposed to design-ing for known creative processes using traditional interac-tion methods.

Future work includes connecting the Camera Keyframingfeature Cochoreo to use embodied methods in the geneticalgorithm. The choreographer would then be able to con-tribute to the initial gene population with their own move-ment data and could create target fitness functions. We arealso investigating options for implementing novelty searchto generate new positions that are maximally different fromwhat the choreographer designs in Cochoreo.

Additional MediaLinks to view videos of movement phrases:

Demonstration Videos of Cochoreo Generative Feature inidanceForms:

Predefined Fitness Function: https://goo.gl/JjqqUc

391

386Proceedings of the Seventh International Conference on Computational Creativity, June 2016

Parameterized Fitness Function: https://goo.gl/AIkwNGDemonstration Videos of Select Choreographers in Final

Cochoreo Study:Participant m: https://goo.gl/gfLcBVParticipant b: https://goo.gl/G3lsHb

ReferencesAlemi, O.; Li, W.; and Pasquier, P. 2015. Affect-expressivemovement generation with factored conditional restrictedboltzmann machines. In Affective Computing and IntelligentInteraction (ACII), 442–448.Blom, L. A. 1982. The Intimate Act of Choreography. Pitts-burgh, Pa: University of Pittsburgh Press.Brand, M., and Hertzmann, A. 2000. Style machines. InThe 27th annual conference on Computer graphics and in-teractive techniques, SIGGRAPH ’00, 183–192. NY, USA:ACM Press/Addison-Wesley Publishing Co.Calvert, T. W.; Welman, C.; Gaudet, S.; Schiphorst, T.; andLee, C. 1991. Composition of multiple figure sequences fordance and animation. The Visual Computer 7(2):114–121.Calvert, T. W.; Bruderlin, A.; Mah, S.; Schiphorst, T.; andWelman, C. 1993. The evolution of an interface for chore-ographers. In Proceedings of the INTERCHI ’93 conferenceon Human factors in computing systems, 115–122. Amster-dam: IOS Press.Carlson, K.; Schiphorst, T.; Cochrane, K.; Phillips, J.;Tsang, H. H.; and Calvert, T. 2015a. Moment by Moment:Creating Movement Sketches with Camera Stillframes. InACM Conference on Creativity and Cognition, C&C ’15,131–140. Glasgow: ACM.Carlson, K.; Tsang, H. H.; Phillips, J.; Schiphorst, T.; andCalvert, T. 2015b. Sketching Movement: Designing Cre-ativity Tools for In-situ, Whole-body Authorship. In The2nd International Workshop on Movement and Computing,MOCO ’15, 68–75. Vancouver: ACM.Carlson, K.; Schiphorst, T.; and Pasquier, P. 2011. Scud-dle: Generating Movement Catalysts for Computer-AidedChoreography. Mexico City, Mexico: ACM Press.Carnegie Mellon University - CMU Graphics Lab - motioncapture library http://mocap.cs.cmu.edu/.Coughlan, T., and Johnson, P. 2009. Understanding Produc-tive, Structural and Longitudinal Interactions in the Designof Tools for Creative Activities. In The Seventh ACM Con-ference on Creativity and Cognition, C&C ’09, 155–164.NY, USA: ACM.Cunteanu, L. 2016. The Power of Co- in ContemporaryDance. Revista-ARTA.Fdili Alaoui, S.; Carlson, K.; and Schiphorst, T. 2014.Choreography As Mediated Through Compositional Toolsfor Movement: Constructing A Historical Perspective. InThe 2014 International Workshop on Movement and Com-puting, MOCO ’14, 1:1–1:6. Paris: ACM.Floreano, D. 2008. Bio-Inspired Artificial Intelligence: The-ories, Methods,and Technologies. Intelligent robotics andautonomous agents. Cambridge, Mass: MIT Press.

Jacob, M; Zook, A; and Magerko, B. 2013. Viewpoints AI:Procedurally Representing and Reasoning about Gestures.In Digital Games Research Association DIGRA 2013.Lapointe, F.-J., and poque, M. 2005. The dancing genomeproject: generation of a human-computer choreography us-ing a genetic algorithm. In The ACM international confer-ence on Multimedia, 555–558. Singapore: ACM.Lapointe, F.-J. 2005. Choreogenetics: the generation ofchoreographic variants through genetic mutations and selec-tion. In The 2005 workshops on Genetic and evolutionarycomputation, 366–369. Washington, D.C.: ACM.Nakazawa, M., and Paezold-Ruehl, A. 2009. DANCING,Dance ANd Choreography: an Intelligent NondeterministicGenerator. In The Fifth Richard Tapia Celebration of Diver-sity in Computing Conference: Intellect, Initiatives, Insight,and Innovations, 30–34. Portland, Oregon: ACM.Russell, S. J., and Norvig, P. 2010. Artificial Intelligence: AModern Approach. N.J: Prentice Hall, 3rd ed edition.Schiphorst, T.; Calvert, T.; Lee, C.; Welman, C.; andGaudet, S. 1990. Tools for interaction with the creativeprocess of composition. In Proceedings of the SIGCHI con-ference on Human factors in computing systems, CHI ’90,167–174. Seattle: ACM.Soga, A.; Umino, B.; Yasuda, T.; and Yokoi, S. 2006.Web3d dance composer: automatic composition of ballet se-quences. In ACM SIGGRAPH, 5 pages. Boston: ACM.Studd, K., and Cox, L. L. 2013. Everybody is a body. Indi-anapolis, IN: Dog Ear Publishing.Yu, T., and Johnson, P. 2003. Tour Jet, Pirouette: DanceChoreographing by Computers. In Genetic and Evolution-ary Computation GECCO 2003. 201–209.

392

387Proceedings of the Seventh International Conference on Computational Creativity, June 2016