We present the first multi-epoch study that includes concurrent mid-infrared and radio interferometry of an oxygen-rich Mira star. We obtained mid-infrared interferometry of S Ori with VLTI/MIDI at four epochs in December 2004, February/March 2005, November 2005, and December 2005. We concurrently observed v=1, J=1-0 (43.1GHz) and v=2, J=1-0 (42.8GHz) SiO maser emission toward S Ori with the VLBA in January, February, and November 2005. The MIDI data are analyzed using self-excited dynamic model atmospheres including molecular layers, complemented by a radiative transfer model of the circumstellar dust shell. The VLBA data are reduced to the spatial structure and kinematics of the maser spots. The modeling of our MIDI data results in phase-dependent continuum photospheric angular diameters of 9.0+/-0.3mas (phase 0.42), 7.9+/-0.1mas (0.55), 9.7+/-0.1mas (1.16), and 9.5+/-0.4mas (1.27). The dust shell can best be modeled with Al2O3 grains alone using phase-dependent inner boundary radii between 1.8 and 2.4 photospheric radii. The dust shell appears to be more compact with larger optical depth near visual minimum (tau_V_~2.5), and more extended with lower optical depth after visual maximum (tau_V_~1.5). The ratios of the 43.1GHz/42.8GHz SiO maser ring radii to the photospheric radii are 2.2+/-0.3/2.1+/-0.2 (phase 0.44), 2.4+/-0.3/2.3+/-0.4 (0.55), and 2.1+/-0.3/1.9+/-0.2 (1.15). The maser spots mark the region of the molecular atmospheric layers shortly outward of the steepest decrease of the mid-infrared model intensity profile. Their velocity structure indicates a radial gas expansion. S Ori shows significant phase-dependences of photospheric radii and dust shell parameters. Al2O3 dust grains and SiO maser spots form at relatively small radii of ~1.8-2.4 photospheric radii. Our results suggest increased mass-loss and dust formation close to the surface near minimum visual phase, when Al2O3 dust grains are co-located with the molecular gas and the SiO maser shells, and a more expanded dust shell after visual maximum. Silicon does not appear to be bound in dust, as our data shows no sign of silicate grains.