Before sending a portrait through a one-bit preview, I wanted to look at the eyes, nose and beard separately from the overall contrast. A continuous-tone image can look crisp at full size while the features become hard to separate after it is reduced.
I used Grok Image 2.0 for a tonal-preparation pass on this portrait reference, then processed the original and the returned image with the same external settings.
The editing request was:
Prepare this same portrait for a grayscale image-preparation comparison. Preserve the man's facial identity, eye shape and gaze, beard, hair, pose, jacket, framing and background. Gently lift the shadows around both eyes and the dark jacket, and reduce the brightest highlights so more facial detail is visible within a narrower tonal range. Keep natural skin texture and wrinkles. Do not smooth the face, alter the features, add outlines, posterize, dither or convert to black-and-white. Return a continuous-tone photographic image with only the tonal adjustment. No text.
The form used Image editing, Standard quality and 2K based on the source image. The result returned as a monochrome continuous-tone image. My prompt was ambiguous about color mode: it called this a grayscale preparation comparison but also said not to convert to black-and-white. Next time I would specify the intended color mode explicitly. The face remained recognizable, but the jacket and some shadow areas looked heavier.
For the lower row, I resized both images to 300 × 225 pixels with Lanczos resampling, converted both to grayscale, and used Pillow's Floyd–Steinberg conversion to one bit. Those steps were performed outside Grok. The displayed previews are enlarged with nearest-neighbor scaling so the dots stay visible.
At that size, I could still locate the eyes and nose in both versions. The edited input looked more defined before dithering, but the one-bit eye areas were still heavily patterned. I could not choose it as a clear improvement from this comparison. The darker jacket also became a larger black area, which changed the balance of the portrait.
For another preparation pass, I would check the same facial regions after conversion before deciding which file to keep. I would also keep the output dimensions and dither settings fixed while comparing inputs. Changing those at the same time would make the source of any improvement harder to identify.
This comparison stops at image preparation. I have no engraved sample from it. For those who prepare photographic portraits, which detail do you judge first in a one-bit preview before moving on to a material sample?