Sieves is an audio-visual series where continuously shifting image scans are interpreted as the parameters for spectral audio signal processing. What is heard as the soundtrack of the video is the video itself interpreted as a sonic effect. In this series, the audio processed through the visual is from field recordings of environmental noise. On screen, rows of pixels are constantly updated to reveal new variations of intensity information evident as grayscale values. This data is utilized to control frequency domain audio filters that enhance or reduce spectral content in recorded sound. The process may be described as a graphically dependent form of spectral subtractive synthesis or spectral noise shaping where analyzed pixel data becomes a tool to sculpt frequency-rich sound sources. Here, grayscale values moving towards black suppress spectral content, while values approaching white reveal it. Graphic form is listenable as new sound shapes are revealed in audio documents.