Generalized Parton Distributions (GPDs) offer a new way to access the quark 
and gluon nucleon structure. We advocate the need to supplement the experimental 
study of deeply virtual Compton scattering (DVCS) i.e. the exclusive 
leptoproduction of a real photon, by its crossed version, timelike Compton 
scattering (TCS) i.e. the exclusive photoproduction of a lepton pair with 
large invariant mass. Both amplitudes factorize with the same GPDs as their 
soft parts and coefficient functions which differ significantly at next to 
leading order in alpha_s. We review recent progress in this domain, 
emphasizing the need to include NLO corrections to any phenomenological 
program to extract GPDs from experimental data.