Unlike hadrons, direct photons are produced in all stages of a nucleus-nucleus collision and therefore test our understanding of the space-time evolution of the produced medium. Of particular interest are so-called thermal photons expected to be produced in a quark-gluon plasma and the subsequent hadron gas. The transverse momentum spectrum of thermal photons carries information about the temperature of the emitting medium. The effect of Doppler blueshift on photons spectra from later and colder stages of a collisions, however, potentially complicates the extraction of the temperature. In this presentation, direct-photon spectra in the range $1 < p_T < 12$ GeV/c from Pb-Pb collisions at $\sqrt{s_{NN}} = 2.76$ TeV will be shown. The results were obtained by measuring $e^+e^-$ pairs from external conversions of photons in the detector material. The measured direct-photon spectra will be compared with predictions from state-of-the-art hydrodynamic models. In the standard hydrodynamical modeling of nucleus-nucleus collisions, thermal photons mostly come from the early hot stage of the collision. As collective hydrodynamic flow needs time to build up, the azimuthal anisotropy of thermal photons quantified with Fourier coefficient $v_2$ is expected to be smaller than the one for hadrons. However, the PHENIX experiment and ALICE experiment observed $v_2$ values of direct-photons similar in magnitude to the pion $v_2$. These unexpected observations constitute the so called "direct-photon flow puzzle" as they challenge the standard hydrodynamic picture of nucleus-nucleus collisions and/or the standard photon emissions rates in the quark-gluon plasma and the hadron gas. We will present the inclusive photon $v_2$ in Pb-Pb collisions at $\sqrt{s_{NN}} = 2.76$ TeV in the range $1 < p_T < 5$ GeV/c and discuss implications for the $v_2$ of direct-photons.