Reducing the emissions of pollutants, and particularly soot particles, from internal combustion engines is one of the greatest challenges faced by car manufacturers. Although modern gasoline direct injection (GDI) engines produce relatively low particulate emissions during steady state operation under near-stoichiometric conditions, they can produce much higher particulate emissions during transients that cause abrupt changes in load, fuel consumption, and the air-to-fuel ratio. Emissions during transients are particularly high when the engine coolant temperature is low, as occurs during engine start-up. Consequently, there is a need to find ways of reducing particulate emissions during load transients. This paper therefore investigates particulate formation during load transients in a single-cylinder GDI engine equipped with an endoscope in the cylinder head. A transient sequence was designed in which the engine load was increased from 4 bar NMEP to a maximum of 12 bar NMEP in 2 s at an engine speed of 2000 rpm. During the transients, the engine’s particulate emissions were measured in terms of particulate number (PN) and images of the combustion process inside the cylinder were captured via the endoscope using a high-speed camera to identify locations where soot formation occurred. Experiments were conducted at a range of coolant temperatures and using different injection strategies to determine how these parameters affect PN emissions. The coolant temperature was found to be the dominant factor governing PN emissions during transients. Luminescence data obtained by analyzing the flame images agreed well with the measured PN emissions during transients. Under all varied parameters in the transients except delayed injection, soot was mainly formed from wall films. For transients with delayed fuel injection, much of the piston film could be avoided but soot formation instead became mixing-dominated. Variation of the air-fuel ratio had little effect on PN emissions during transients. At all coolant temperatures, PN emissions were lowest when using a split injection strategy but delaying the injection timing increased PN emissions even though the endoscope images suggested a lower frequency of diffusion flame formation. No conditions were found under which the PN emissions during transients with low coolant temperatures could be reduced to levels comparable to those seen with warm coolant.