The European Centre for Medium-range Weather Forecasts is increasingly using cloud and precipitation-affected radiances in its numerical weather prediction system. The intention is to improve initial conditions of temperature and wind in areas that are otherwise difficult to observe, leading to better forecasts, as well as simply to improve the analysis of moisture, cloud and precipitation. Microwave imager radiances (e.g. SSMIS, AMSR-E) are assimilated operationally in clear, cloudy and precipitating conditions, using an 'all-sky' approach. This is being extended to the lower-tropospheric channels of AMSU-A, benefitting temperature analyses and forecasts in the tropics, though it is difficult to get improvements elsewhere. One concern is the simulation of scattering radiative-transfer in falling snow, but there are promising results when switching from the soft Mie-sphere to the discrete dipole approximation. In the infrared, temperature-sensitive advanced sounder radiances (IASI and AIRS) are assimilated operationally in clear and completely overcast situations, but cloud information is discarded. A unified approach is in development that will constrain both cloud and temperature in the analyses and allow assimilation in partially-cloudy situations.