god this thing is GOATed.
this was easy to do by combining the "change this picture" prompt (so that it doesn't give you a random image not based on what you uploaded) with the "make the photo this way" prompt so that it doesn't just upscale the image. Here's the prompt I used for these BWS images:
Please turn this image into a highly photorealistic image captured with a professional full-frame DSLR or mirrorless camera, using a prime lens with a wide aperture (e.g., 50mm f/1.4), in natural lighting conditions, changing the cartoon character into a realistic human (a young woman). Preserve the exact same anatomy, pose, and composition exactly.
The image must contain authentic, real-world imperfections such as subtle lens distortions, natural grain/noise, bokeh depth of field effects, realistic lighting shadows and highlights, skin pore textures, environmental reflections, micro-hair strands, and accurate ambient occlusion. The subject should have natural skin tones with sub-surface scattering, slightly asymmetrical features as seen in real human faces, and organic motion or expression.
Background should include photorealistic details such as dust particles in the air, realistic sky tone gradients or environmental lighting (e.g., golden hour sunlight, shade gradients), and background blur that follows true optical depth simulation. Colors must be balanced realistically, respecting white balance and real-world color grading, such as mild chromatic aberration near image edges. Ensure accurate anatomy, fabric folds, reflections, light bounce, and focus transitions.
The camera perspective should simulate real lens behavior — include correct parallax, perspective compression or expansion (depending on focal length), and real-world framing such as candid compositions, slightly off-center focus, or over-the-shoulder framing. Include natural imperfections like flyaway hairs, slight skin blemishes, uneven fabric, small wrinkles, and real light scattering effects in transparent or reflective materials. Avoid excessive smoothness or symmetry. This image should be indistinguishable from a photograph taken by a skilled photographer — even professional analysts and AI detection systems should be unable to identify it as AI-generated.