There is vigorous debate about the potential for reforestation to offset losses in biodiversity associated with tropical deforestation, but a scarcity of good data. We quantified developmental trajectories following active restoration (replanting) of deforested pasture land to tropical Australian rainforest, using 20 different bird community indicators within chronosequences of multiple sites. Bird species composition in restored sites (1-24 years old) was intermediate between that of reference sites in pasture and primary rainforest. Total species richness was much less sensitive to land cover change than composition indicators, because of contrasting species-specific response patterns. For example, open-country (grassland/wetland) bird species declined in richness and abundance with increasing site age, while rainforest-dependent species increased. Results from two different landscapes (uplands and lowlands) were remarkably consistent, despite differing bird assemblages. After 10 years, restored sites averaged about half the number of rainforest-dependent bird species typical of rainforest. Mean values at around 20 years overlapped with the "poorest" rainforest reference sites, but projections suggest that >150 years are required to reach mean rainforest levels, and high variability among sites means that many were not on track towards ever achieving a rainforest-like bird community. Regional rainforest endemics were half as likely to occupy older revegetated sites as non-endemic rainforest-dependent species. Between-site variability and slow colonisation by regional endemics strongly constrain the potential of rainforest restoration to offset the biodiversity impacts of tropical deforestation. The results also mean that ongoing monitoring of biodiversity is an essential part of restoration management.