ENFPDEV

This & That: complex edits, asked simply. Draw and talk at the same time. AI image and 3D tools get you 90% there. The last 10% is details, and words can't point. With This & That you circle or sketch on the picture (or on a 3D model) while you speak: "make this gem this shape", "remove this, make this red, and add a little star here". Meta's models turn the strokes and the voice into one instruction, and only what you marked changes. - Image mode: shape sketches, references, and several edits in one breath. Everything outside the marks stays pixel-identical: text-only edits changed 10.4% of the untouched pixels, This & That 0.6%. - 3D mode (beta): reshape a low-poly model by drawing (lengthen the blade, bend it into a katana, set a gem) and bake detail edits into a high-quality model's texture. Measured on 6 hard cases: 124 typed words vs about 30 spoken. "Third row, fifth from the left" changed the wrong macaron; one tap got it right. Try it live: https://this-and-that.enfpdev.workers.dev (passcode: willow-ember-87). Chrome recommended; a tablet with a stylus works best.