|
← 判断力与美学 ← Judgment & Aesthetics
SAE 判断力与美学 · 余项之美
SAE Judgment & Aesthetics · Beauty of the Remainder
2026-04-22

Andrei Jay:视频反馈作为余项的乐器制造

Andrei Jay: Building Instruments for the Remainder of the Loop

Han Qin (秦汉)

视频反馈是这样一件事:把摄像头对准它自己输出的屏幕,或者把帧缓冲区的输出重新送回输入。每一帧都是前一帧的残余——被位移、被变形、被叠加之后的沉积物。这不是比喻意义上的余项,而是物理结构上的:帧N是帧N-1经过非线性变换后的剩余。数学家出身的视频艺术家Andrei Jay(网络身份ex-zee-ex)围绕这个结构建造了一整个开源乐器生态系统。他把它叫做Video Synthesis Ecosphere——一个由Raspberry Pi驱动的、可互换固件/软件镜像组成的平台,每个镜像是一件不同的视觉乐器。

这个生态系统包含多件乐器:WAAAVE_POOL是帧缓冲延迟/视频反馈生成器,操作者通过MIDI控制器实时塑造反馈回路的参数——延迟时间、位移量、色彩空间的扭转;AUTO_WAAAVE是它的音频反应版本,声音信号驱动视觉反馈的变异;SPECTRAL_MESH是一个受扫描处理启发的视频再合成器,用一组混沌视频振荡器实时扭曲和变形输入信号;PHOSPHORM是为X-Y示波器显示优化的视听合成器。Jay用了一个精确的类比:整个系统像任天堂红白机,硬件是固定的(一台Raspberry Pi 3b、一个USB摄像头、一个MIDI控制器),每件乐器是一盘不同的卡带。你换一张SD卡,就换了一种生成视觉余项的方式。

为什么这是余项之美而不是已构之美?艺术界有"视频艺术"这个类别,有"模拟视频合成"这个子类,有Nam June Paik到Steina Vasulka的经典谱系。这些都是已构——它们的凿构循环已经完成,被写进了美术馆的永久收藏和教科书。Jay做的事情不在这个谱系里。他不是在做视频艺术作品,他在做生成视频余项的工具。区别是根本性的:一件作品是凿构循环的产物,而一件乐器是余项持续生产的场所。WAAAVE_POOL不产出固定的作品,它产出的是一个实时运行的反馈空间,操作者在其中与非线性动力学的混沌结构持续互动。每一次演出、每一次使用都是不可重复的——不是因为随机性,而是因为反馈系统对初始条件的敏感依赖使得每一条轨迹都是独一无二的余项。

更关键的是Jay选择的介质:消费级硬件和开源代码。他不在画廊里展示,他在Echo Park Film Center办工作坊。他的代码在GitHub上,任何人都可以fork。社区已经在自发生长:有人把WAAAVE_POOL移植到了Raspberry Pi 5,有人写了详细的使用指南,有人在此基础上开发新的乐器。这不是一个艺术家的个人项目,这是一个乐器制造传统的开端——就像Buchla和Serge定义了模块化音频合成器的语法,Jay正在为视频反馈合成器定义一套消费级、开源的语法。但这个定义还在进行中,还没有被任何一个领域(视觉艺术?音乐?工具制造?开源文化?)完整认领。

现在看到这件事比以后看到更重要。视频合成正在经历一次从昂贵的模拟硬件到廉价数字平台的迁移,AI生成视觉正在从另一个方向侵蚀这个领域。Jay的生态系统恰好处于两者之间的缝隙:它是数字的但不是AI的,它是廉价的但不是一次性的,它是工具但它的操作本身就是美学行为。一旦"开源视频合成器"或"Raspberry Pi视觉乐器"成为一个稳定的类别,Jay的工作就会变成已构——变成这个类别的起源故事。但现在,它还是余项。反馈回路还在跑,帧还在累积,混沌还没有被命名。

videosynthesisecosphere.com ↗

Video feedback works like this: point a camera at the screen displaying its own output, or feed a framebuffer's output back into its input. Every frame is the residue of the previous one — displaced, deformed, accumulated. This is not remainder in a metaphorical sense but in a structural one: frame N is what remains of frame N-1 after a nonlinear transformation. Andrei Jay, a mathematician turned video artist (known online as ex-zee-ex), has built an entire open-source instrument ecosystem around this structure. He calls it the Video Synthesis Ecosphere — a platform of interchangeable firmware/software images running on Raspberry Pi, where each image is a different visual instrument.

The ecosystem contains several instruments: WAAAVE_POOL is a framebuffer delay and video feedback generator, where the operator shapes the feedback loop's parameters in real time via MIDI controller — delay duration, displacement magnitude, color-space distortion. AUTO_WAAAVE is its audio-reactive sibling: sound signals drive the mutations of visual feedback. SPECTRAL_MESH is a scan-processing-inspired video resynthesizer that uses a set of chaotic video oscillators to warp and deform input signals in real time. PHOSPHORM is an audiovisual synthesizer optimized for X-Y oscilloscope display. Jay uses a precise analogy: the whole system is like a Nintendo Entertainment System. The hardware is fixed (a Raspberry Pi 3b, a USB camera, a MIDI controller); each instrument is a different cartridge. Swap the SD card, and you swap the method of generating visual remainder.

Why is this beauty-of-the-remainder rather than beauty-of-the-construct? The art world has "video art" as a category, "analog video synthesis" as a subcategory, and a canonical lineage from Nam June Paik to Steina Vasulka. These are all already-construct — their chisel-construct cycles are complete, written into museum permanent collections and textbooks. What Jay does sits outside this lineage. He is not making video art works; he is making tools that generate video remainder. The distinction is fundamental: a work is the product of a chisel-construct cycle, while an instrument is a site where remainder is continuously produced. WAAAVE_POOL does not output fixed works. It outputs a real-time feedback space in which the operator continuously interacts with the chaotic structures of nonlinear dynamics. Every performance, every session is unrepeatable — not because of randomness, but because feedback systems' sensitive dependence on initial conditions makes every trajectory a unique remainder.

Equally critical is Jay's choice of medium: consumer-grade hardware and open-source code. He does not exhibit in galleries; he runs workshops at Echo Park Film Center. His code lives on GitHub; anyone can fork it. The community is already growing on its own: someone has ported WAAAVE_POOL to Raspberry Pi 5, others have written detailed guides, still others are developing new instruments on top of the platform. This is not one artist's personal project — it is the beginning of an instrument-building tradition. Just as Buchla and Serge defined the grammar of modular audio synthesizers, Jay is defining a consumer-grade, open-source grammar for video feedback synthesizers. But that definition is still in progress, not yet fully claimed by any single domain — visual art? music? toolmaking? open-source culture?

Seeing this now matters more than seeing it later. Video synthesis is undergoing a migration from expensive analog hardware to cheap digital platforms, while AI-generated visuals are encroaching from the other direction. Jay's ecosystem occupies exactly the gap between both: it is digital but not AI, cheap but not disposable, a tool whose operation is itself an aesthetic act. Once "open-source video synthesizer" or "Raspberry Pi visual instrument" becomes a stable category, Jay's work will become construct — the origin story of that category. But right now, it is still remainder. The feedback loop is still running, frames are still accumulating, and the chaos has not yet been named.

videosynthesisecosphere.com ↗