DTF halftone knockout uses the distance between a pixel’s RGB color and the selected color. An exact match becomes transparent. Within the tolerance region, the remaining alpha increases gradually with distance; beyond that region, the original alpha is retained. This continuous falloff is then converted into dot coverage. The tolerance control is not a subject-selection tool, a perceptual color match or an automatic background remover.
Minimum dot diameter is tied to physical units. At 300 DPI, 0.3 mm corresponds to roughly 3.54 pixels; at 600 DPI it is roughly 7.09 pixels. A mark below that threshold is omitted. Increasing DPI while leaving the pixel grid unchanged therefore removes more small marks. That is expected: the same number of pixels now represents a smaller physical feature. DPI metadata alone does not add detail to the image.
Six shapes remain available in DTF mode: circle, square, diamond, line, ellipse and cross. The minimum-size rule uses the narrow dimension for an ellipse, line or cross, so a long mark does not pass solely because it is long. The cell average determines the nominal mark size, and the source alpha mask prevents the dot from painting into fully transparent source pixels. Retained opaque pixels keep their original RGB color.
PNG is the clearest transfer format for this workflow. Its pHYs chunk records the selected pixels-per-meter value corresponding to your DPI. Some receiving software may ignore this metadata, so verify placement size after importing. SVG is also available, but intact opaque artwork can contain many colors: the DTF halftone generator stores same-color horizontal pixel runs as vector paths. Above 250,000 runs, it asks you to confirm before building the file.