Dream Machine: Live Performance Prototype
This week, I focused on turning Dream Machine from a sequence of pre-authored scenes into a more playable live performance system. A large part of the work happened simultaneously across music composition, Unreal Engine interaction design, lighting, scene transitions, and physical MIDI control. Instead of treating audio, camera, and environment as separate elements, I started connecting them as parts of one continuous performance.
这周我主要把 Dream Machine 从一组预先编排好的场景，逐渐转变成一个更具有“现场演奏感”的实时表演系统。大量工作同时发生在音乐创作、Unreal Engine 交互设计、灯光、场景切换以及实体 MIDI 控制器之间。相比把声音、镜头和环境当成彼此独立的元素，我开始尝试把它们连接成一个连续的表演系统。
On the music side, I completed most of the composition and scene-mix preparation in Ableton Live. The full experience now follows a continuous musical arc across three memories: Office, Subway, and Station. All of the music is built around 112 BPM in 4/4, so Unreal can use the same musical clock to synchronize transitions and visual events.
在音乐部分，我基本完成了 Ableton Live 中的作曲和 Scene Mix 准备。目前整个体验已经形成了一条跨越三个记忆场景的连续音乐线：Office、Subway 和 Station。所有音乐统一建立在 112 BPM、4/4 拍的基础上，因此 Unreal 可以使用同一个音乐时钟来同步场景切换和视觉事件。
Rather than rebuilding the Ableton project from individual stems inside Unreal, I exported complete musical scenes and created a runtime system that can preload and crossfade between them on musical boundaries. This makes the transitions more reliable and allows the music to remain continuous even while the environment changes.
我没有选择把 Ableton 里的 Bass、Synth、Motif 等多个 stem 再次拆开并在 Unreal 中重新构建，而是直接导出完整的音乐 Scene，并建立了一套运行时播放系统。系统可以提前加载下一段音乐，并在拍点或小节边界进行交叉淡化切换。这样不仅让转场更加稳定，也让音乐可以在环境发生变化时保持连续。
The Office scene went through several interaction revisions. The opening now reacts to the relationship between camera orbit speed and framing instead of immediately progressing after a single knob movement. The player gradually discovers that slower observation and wider framing can unlock the next musical and visual stage.
Office 场景这周经历了多次交互调整。现在开场不再因为玩家随便转一下旋钮就立刻推进，而是根据摄像机环绕速度和镜头取景范围之间的关系来决定是否解锁下一阶段。玩家需要逐渐发现：更慢的观察速度和更广的画面范围，会推动音乐和视觉进入下一层状态。
I also introduced a manual cue system. Previously, once a transition condition was satisfied, the next audio scene would trigger automatically. This often made the pacing feel too mechanical. Now, satisfying a condition only makes the next transition ready.
我还加入了一套手动 Cue 系统。之前只要交互条件满足，下一段音乐就会自动触发，这会让整体节奏显得有些机械。现在，条件满足只代表“下一段已经准备好”，而不会直接执行。
The interface now uses a small status light to communicate the cue state. Yellow means the next transition is not yet available, flashing green means it is ready, and blue means it has already been cued. The performer presses Pad 10 to confirm the next transition, allowing the timing to be controlled more intentionally during live performance.
现在界面会通过一个小型状态灯来提示 Cue 状态。黄色表示当前还不能切换，绿色闪烁表示下一段已经准备好，蓝色则表示已经完成 Cue、正在等待音乐边界执行。表演者可以按下 Pad 10 来确认下一段，这让现场表演中的节奏控制更加主动。
UI Before
UI Now
I also simplified the HUD. Essential controller instructions and cue information are placed more clearly, while technical information is pushed into a quieter gray background layer. This makes the interface feel less like a debugging screen and more like part of the performance.
我还重新整理了 HUD。最重要的控制器说明和 Cue 信息会以更清晰的视觉层级呈现，而 BPM、Bar、镜头参数等技术信息则被弱化成较浅的灰色背景信息。这样界面不再像开发调试工具，而更像表演系统本身的一部分。
before
The Subway scene changed significantly as well. I replaced the original blockout train with a more detailed modular subway asset while preserving the existing interaction systems. The player can still control TimeFlow, train vibration, camera parameters, and moving exterior lights.
Subway 场景这周也发生了很大的变化。我用一套更完整的模块化地铁资产替换了原先的 blockout 车厢，同时保留了原有的交互系统。玩家依然可以控制 TimeFlow、列车震动、摄像机参数以及窗外运动光线。
The exterior lighting now enters from both sides of the carriage and rotates at different speeds and phases. Knob 5 controls the moving-light speed, from almost stationary to extremely fast, creating a much stronger sense of motion.
现在外部光线会从车厢左右两侧同时进入，并以不同的速度和相位进行旋转。Knob 5 用来控制这些运动光线的速度，从几乎静止到非常快速，从而制造更明显的列车运动感。
I also added rotating scenery systems outside the windows, inspired by revolving stage sets in theatre. Different cues can rotate different exterior environments into view, creating a deliberately theatrical and dreamlike feeling of changing scenery outside the train.
我还在车窗两侧加入了类似舞台剧转台的场景切换系统。不同的 Cue 可以让不同的外部场景旋转到车窗外，形成一种有意保留“舞台装置感”的梦境式空间切换。
另一个实验是地铁中的物理物体洪流。在某个随机选择的 Cue 中，数百个小型物理物体会从摄像机后方涌入车厢，并逐渐堆积在地面上。下一次 Cue 时，车门打开，大部分物体会通过真实物理模拟被推出车外，同时仍然会有一部分残留在车厢内部。
These objects use actual physics and respond to the train vibration, so the train itself feels more physically alive. This part of the system is still being optimized for performance.
这些物体使用真实物理模拟，并会受到列车震动的影响，因此整个车厢在视觉上更具有物理生命感。目前这一部分仍然在继续进行性能优化。
The Subway lighting also became more theatrical. Instead of repeatedly turning the entire carriage completely black, the music now modulates the light between bright and dim states. During the tunnel section, external neon and laser-like colored lights become the main source of illumination inside the carriage.
地铁灯光的方向也逐渐变得更舞台化。相比让整个车厢随着鼓点反复完全变黑，现在音乐主要控制灯光在明亮和较暗状态之间变化。在隧道部分，来自车窗外部的霓虹和类似激光的彩色光线会成为车厢内部的主要照明来源。
The third memory, Station, now works differently from the first two scenes. Instead of playing music in a fixed sequence, three synchronized music beds blend continuously according to the camera position and viewing direction.
第三个记忆场景 Station 的逻辑与前两个场景不同。它不再按照固定顺序切换音乐，而是让三段同步的音乐根据摄像机的位置和观看方向进行连续混合。
The Station camera is also fully free after the arrival sequence. The player can rotate, change distance, and explore the space, while the sound field changes continuously around them.
Station 在进入完成后，摄像机也会完全释放为自由探索状态。玩家可以自由旋转视角、改变距离并探索空间，而音乐场会根据这些移动持续发生变化。
The environmental sound system is also prepared for spatial textures such as wind, leaves, insects, distant voices, room tone, and other memory-like sounds. These sounds are independent from the 112 BPM musical structure and can exist as free-running spatial MetaSounds.
环境声音系统也已经为风声、树叶、昆虫、远处人声、Room Tone 以及其他类似记忆纹理的声音预留了空间。这些声音不会依赖 112 BPM 的音乐结构，而是作为自由运行的空间 MetaSound 存在。
The next major step will be character animation. I want repetitive actions such as typing and dancing to synchronize with the same 112 BPM musical clock, while one-shot movements such as looking up, falling, standing, or sitting begin on meaningful musical beats without being unnaturally stretched.
下一步的重点将是人物动画。我希望 Typing、Dance 等重复动作能够和同一个 112 BPM 音乐时钟同步，而 Look Up、Falling、Stand Up、Sit 等一次性动作则在合适的 Beat 或 Bar 上开始，同时保持动作本身自然，不进行过度拉伸。