Revealing the physicochemical characteristics of exhaled particles is essential for understanding and efficiently mitigating the airborne spread of contagious human illnesses. Among the most pivotal factors, the number size distribution of emitted particles plays a crucial role when considering atmospheric dispersion. This study focuses on submicron particles emitted during speaking, with particular attention on the changes over time. Moreover, the real-world (source control) efficiency of three types of commonly used facemasks (FFP2, surgical and 2-layer cotton mask) under in vivo conditions was studied. A specially designed cabin ensured a controlled environment, where a set of experiments was conducted on 28 participants. Our findings revealed no substantial variability in the number size distribution among different individuals and pitches. However, the quantity of emitted particles varied significantly among individuals, with differences reaching nearly two orders of magnitude. Additionally, the emitted number of particles strongly depended on the speaking volume, decreasing as speech volume was reduced. Submicron particles originating from the lungs and upper airways exhibited a consistent bimodal pattern, with peaks around 300 nm and below 100 nm. FFP2 and surgery masks worn by the subjects demonstrated robust performance in real-world conditions characterized by 80% source control efficiency even for the smallest particle size ranges tested. At the same time, textile masks yielded less favourable results of 50-60% source control efficiency.