We present the pipeline processing (for photometry and spectrophotometry) primarily based on the comparison of the observation with a prediction from the spectral energy distribution and total flux of the source and a set of parameters modelling the instrument properties observation-to-observation. Experiments on G white light calibration yield residuals well below one mmag for most of the cases, but for extreme colour stars. For blue and red spectra, tests up to now yield residuals significantly smaller than the observational noise, except for sources with very strong spectral features, which present flux residuals of 2–3% in the spectral lines (at G = 15). Work continues to account for all instrumental effects, to add complexity to the instrument model and to improve the processing pipeline.