The Star Imager performance under various ground-based operating conditions is discussed. The precision is generally found to be in the range of 1 arcsecond 1 pointing and 5 to 8 times larger for the measured roll about the boresight. Even though a carefully planned and conducted test series has been performed, the systematic errors and test-setup related biases and noise sources seem to dominate the measurements. The instrument is then expected to show a better performance in space. Particular emphasis has been directed towards the performance of the software when exposed to images acquired during attitude manoeuvres, i.e. motion smeared images. In general it is found that the high precision software performs well for attitude change rates up to 1/3 deg/sec and with little degradation. But the performance rapidly decreases at higher rates. However, by adapting the software, the operation of the Star Imager may be extended to 5-6 deg/sec, as demonstrated by the "Thunderstorm III" flight data, or by decreasing the integration time, as for the ASTRID II micro satellite version of the instrument, expected to operate at rates up to 18 degrees per second.