Such a question, need to make pixel art 2D 64x64 from 3D models/photo as a need to process pictures so they were as similar to the drawings. With a thick black outline, saw that it is necessary to make a vector image with flat colors, but still looks too blurry =(
The example you linked to one of the comments is far away from a photo. It looks a well crafted vector drawing with well separated details. Assuming you accept a more reasonable size than 64 x 64 it can be pixelized easily and without losing everything:
It got in Photoshop
Image > Image size > Resample to 120 x 160 pixels with Bicubic resampling
Filter > Sharpen > Sharpen edges
Image > Mode > Indexed color > 16 colors, use adaptive palette and primary colors
Step 3 is shown in the previous image.
The number of colors is set to 16 because old computers and games didn't have more at a time - having limited palette became characteristic to pixel art. You can leave out the whole step 3 if it can have all RGB colors.
The lips are no more red due the small color palette. To add red and at the same time to keep the color palette limited return the image mode to RGB and tweak one or a few pixels manually. In the next image the lips are selected with non-antialiased zero feathering polygonal lasso and Image > Adjustments > Hue/Saturation, hue shift towards red (no saturation nor brightness adjustment) is applied:
Now the lips catch more attention, you may want to take the pencil tool and remove also some dark spots:
This leads easily to an endless loop. It's difficult to stop, but one must stop if he's going to get the job done in a reasonable time. In addition the image was already optimized with a math criteria and changing something breaks that balance, so it's not at all clear that my edits made it look better when seen by someone else. A real artist would be needed to make proper tweaks. If you are one, then there's no problem.
It's useless to add a couple of new reds for the lips to the indexing palette and to try to convert the original downscaled RGB version to indexed mode with the augmented palette. The new reds will appear everywhere and lose totally their impact.
As said, the example does not resemble a photo. The photos are generally much more complex and the separation between different features is often subtle. The method above is useless for photos. But converting a photo to pixel art can still succeed in Photoshop, GIMP etc... if a human separates different parts and converts them one by one. An example:
This old-timer is pixelized to 16 colors. Low contrast in the face and the limited palette renders his face to ugly mess. Doubling the amount of colors doesn't help because the whole image is full of subtle color differences. There's no easy way to budget more colors for the face, which is the most important part in many cases.
In the next version the head is converted separately and pasted back:
It also needed some minor tweaks with the 1 px pencil to remove a couple of bad spots. Separating the parts, deleting all non-essential ones, pixelizing and recombining is very time consuming when compared to more intelligent conversion methods. For a good non-messy result one must separate all wanted items and delete the background. Separating only the head made it recognizable, but everything else is still far too busy.