Performance Video 1 (Classical style; ~12mins 1080p MP4)
Screen capture of live music generation with inset platform footage, classical style using piano and synthetic choir (pro-grammed to sing the neighbourhoods of Bristol). The segment at 5:00-7:00 (2 minutes) illustrates how the music respond to the transition from a quiet platform to, upon train arrival, a busy platform.

Listening Guide / Key Elements:

• Crowd density affects overall instrumentation, tempo and dynamic (quiet = soft and slow; busy = fast and heavy). The choir is silent during quiet periods.
• Composer dictates harmonic language, but crowd determine both pitch and dynamic of individual notes.
• Broadly tonal palette based on a randomised tonic-predominant-dominan chord progression
  (voiced using an adaptation of Pärt's Tintinnabuli technique)
• Individual persons clustered into 5 Groups, to build and voice the chord
  (shown on the bottom edge of the detection window)
• Groups are weighted by size and proximity, such that the motion of individuals shifts and rebalances groups, appreciable in the music.
• The presence of specific objects (e.g. bicylces) triggers more dissonant textures, prompting the additions of sevenths and ninths (e.g. at 8:00).
• Synthetic choir sings Bristol neighbourhoods ("Clifton", "Totterdown", "Frenchay", etc.), using concatenative synthesis to sing a prepared script to live notes.

Performance Video 2 (Trip Hop style; ~12mins 1080p MP4)
Screen capture of live music generation with inset platform footage, trip-hop style using synthesisers and sampled Rising Voices choir (for both voice and drum sounds). The segment at 5:00-7:00 (2 minutes) illustrates how the music respond to the transition from a quiet platform to, upon train arrival, a busy platform.

Listening Guide / Key Elements:

• Developed using slightly simpler mapping techniques, easier for a public audience to appreciate.
• Crowd density affects overall instrumentation, mix balance, and drum layering/programming.
  • Busier periods feature more densely layered drum parts.
  • Synth bass only present when Suitcase is visible ("people can walk off with the bass").
  • Drum loop (Amen break) is more likely to appear in the presence of trains or bicycles.
  • More people mean more choral parts (from none to three).
  • A slow, atmospheric, filter-swept synth pad plays during quiet periods on the platform.
• Choir and drum sounds recorded from Rising Voices (Bristol's Recovery Choir)
• Multi-sampled "ah" and "ee" sounds for pitched choir voice.
• Additional vocal noises for percussion (all drums except Amen break).
• Uses simplified I-IV-V harmony map, density mapped to crowd.
• Some non-deterministic processes used to avoid repetitive textures during sustained extremes (such as deserted platforms or rush hour)

Early Technical Demo (annotated initial experiments; 2:34)
Annotated screen capture of early experiments with basic music mappings of people locations to notes (pitch and rhythm) and percussion density to crowd density. Annotations are provided to explain the system and process.

Listening Guide / Key Elements:

• Early experimentation with basic mapping techniques to test technology and explore expressive possibilities.
• No specific artistic goal or musical aeshetic applied.
• Developed as a proof-of-concept and technical demo for funders and collaborators (i.e. BBC, GWR).
• Concepts and techniques are illustrated through a visual commentary.
• Linear, geometric map of individual crowd members to note pitch and rhythm, using cartesian coordinates.
• Horizonal offset (X) maps to part and pitch, broadly proportional to pitch height (left-right > low-high), but split into five vertical bands mapped to separate parts (for 5-note polyphony). Position-to-pitch map is linear, but quantised to notes in a defined scale to improve consonance.
• Vertical offset (Y) maps to rhytmic offset within a repeating pattern. Mapping technique exploits Manhattan's facility to address future events on a linear timeline, supporting coherent music responses using the periodicity of bars and repeated phrases.
• Tuned percussion selected to suit the resulting aeshetic; the partially-harmonic/inharmonic timbre aligns with partially-consonant/dissonant note texture.
• Crowd density affects note density and drum layering/programming (e.g. 2:00):
  • Number of people can be linearly mapped to musical parameters, or classified into descriptive categories (empty, quiet, busy, etc.) based on length of the list of detections (shown in Channel 1).
  • Busier periods feature more densely layered drum parts.
  • Drum loops play during busier periods.
  • Loop is randomly recycled (cut up and resequenced) on-the-fly at peak periods.
  • An acoustic bass part selects randomly from C, G and A, as target notes using prescribed intervals to add one or two subsequent notes in the bar, depending on crowd density, to act as passing notes.