The level-1 data of SDO/HMI have been inverted with the UNNOFIT inversion code (Bommier et al., 2007, A&A, 464, 323), which differs from VFISV about the magnetic filling factor modeling. More realistic field inclinations are obtained outside the active region. The spatial resolution seems to be also better. UNNOFIT is enabled for the taking into account of gradients of radial velocity, responsible for asymmetry of the Stokes profiles (Molodij et al., 2011, A&A, 531, A139). The ambiguity has been solved with the ME0 code of Metcalf, Leka, Barnes & Crouch. We present the movie of 4 hours of observation, the flare occurring at middle. Two current ribbons of opposite polarity are visible along the magnetic neutral line, in the vertical density current map. The negative one strengthens and radially expands from the flare center at the eruption moment. A similar current ribbon pair is visible in a 3D line-tied zero-beta resistive MHD flare simulation with the OHM code (Aulanier et al. , 2012, A&A, 543, A110). The two ribbons part from the flare center during the flare, similarly to what observed on 2011 February 15 at 02:00 with SDO/HMI.