In deeply virtual Compton Compton scattering (DVCS), the number of independent amplitudes is known, namely five if the initial and final photons are virtual. This number reduces to three if either one is real, which is the case in for instance the experiments done at JLab. The most economical way to describe the amplitude for this process uses a second-rank tensor that gives the amplitudes by contraction with the photon wave functions. Methods to construct this Compton tensor will be discussed and implemented in a simple solvable model. It will turn out that the Compton form factors behave for large virtuality Q^2 of the initial photon like log(Q^2)/Q^2.