MI 445 Project 1

Originally, Parry Hero (https://palash-mathur.itch.io/parry-hero) analyzed music in real time while the song was playing. Unity’s spectrum data was checked every frame, and bass, mid, or treble spikes would immediately trigger obstacle spawning. This worked for early testing, but it made the system very reactive and gave us little control over what was coming next in the level.

The new system analyzes the entire song first using an offline FFT. Instead of only looking for raw frequency spikes, it stores sustained activity, how much the signal changes compared to the previous few seconds, and specific musical events. That information is saved into SongAnalysisData.

A separate LevelGenerator then reads that analysis in 0.5-second windows and translates it into gameplay intent for three simultaneous routes. Beat controls bhop-focused movement, Mid controls slide-focused movement, and High controls aerial combat, parries, and kills. This separates the music analysis from the actual level generation, which gives us much more control over how the song influences gameplay.

The graphs are a visual representation of that process. The top section shows the audio analysis for Beat, Mid, and High. The height and brightness of the colored blocks represent sustained activity in each part of the music, while the white vertical bars represent detected musical spikes or events that stand out from the recent average. The second section shows how that information is translated into gameplay intent. Darker sections represent safer or simpler gameplay, while brighter sections represent more active or complex bhop, slide, or aerial-combat sections.

Overall, the old system was essentially music event → immediate obstacle, while the new system is song analysis → stored music data → gameplay decisions → level generation. This gives us a better foundation for eventually generating, checking, and validating an entire playable level before the player starts.