Full-disc observations of the Sun in the H{alpha} line provide information about the solar chromosphere, and in particular, about the filaments, which are dark and elongated features that lie along magnetic field polarity-inversion lines. This makes them important for studies of solar magnetism. Because full-disc H{alpha} observations have been performed at various sites since the second half of the 19th century, with regular photographic data having started at the beginning of the 20th century, they are an invaluable source of information on past solar magnetism. We derive accurate information about filaments from historical and modern full-disc H{alpha} observations. We consistently processed observations from 15 H{alpha} archives spanning 1909-2022. The analysed datasets include long-running ones such as those from Meudon and Kodaikanal, but also previously unexplored datasets such as those from Arcetri, Boulder, Larissa, and Upice. Our data processing includes photometric calibration of the data stored on photographic plates, the compensation for limb-darkening, and the orientation of the data to align solar north at the top of the images. We also constructed Carrington maps from the calibrated H{alpha} images. We find that filament areas, similar to plage areas in CaII K data, are affected by the bandwidth of the observation. Thus, a cross calibration of the filament areas derived from different archives is needed. We produced a composite of filament areas from individual archives by scaling all of them to the Meudon series. Our composite butterfly diagram very distinctly shows the common features of filament evolution, that is, the poleward migration as well as a decrease in the mean latitude of filaments as the cycle progresses. We also find that during activity maxima, filaments cover ~1% of the solar surface on average. The change in the amplitude of cycles in filament areas is weaker than in sunspot and plage areas. Analysis of H{alpha} data for archives with contemporaneous CaII K observations allowed us to identify and verify archive inconsistencies, which also have implications for reconstructions of past solar magnetism and irradiance from CaII K data.