You need 4K plates for a long, slow push-in. Layers come back at about 2K. Scaled up for a 4K composition, edges soften on screen. Upscale first, or paint the plates at full size.
The scene has dozens of moving parts. One split returns at most 16 layers, so small parts get grouped. Split crops of the busy areas separately, or combine passes.
You want kinetic type from the poster text. Text comes back as a picture of the letters, not a type layer. Hide it and retype the words in After Effects to animate them character by character.
The source is video, not a still. Separating a subject across moving footage is rotoscoping inside After Effects. An image splitter works on single frames.
You still have the designer's layered file. Use it. Nothing rebuilt from a flat export is better than the original layers.
Hidden parts are usually drawn in, as with the cup above, but not always: on our character sample the neck, hands and ankles were not separated and had to be kept as an extra layer. Check each part at full size before you rig it.