The aim of the present study was to validate a food photograph album (FPA) as a tool to visually estimate food amounts, and to compare this estimation with that attained through the use of measuring cups (MC) and food models (FM). We tested 163 foods over fifteen sessions (thirty subjects/session; 10–12 foods presented in two portion sizes, 20–24 plates/session). In each session, subjects estimated food amounts with the assistance of FPA, MC and FM. We compared (by portion and method) the mean estimated weight and the mean real weight. We also compared the percentage error estimation for each portion, and the mean food percentage error estimation between methods. In addition, we determined the percentage error estimation of each method. We included 463 adolescents from three public high schools (mean age 17·1 (sd 1·2) years, 61·8 % females). All foods were assessed using FPA, 53·4 % of foods were assessed using MC, and FM was used for 18·4 % of foods. The mean estimated weight with all methods was statistically different compared with the mean real weight for almost all foods. However, a lower percentage error estimation was observed using FPA (2·3 v. 56·9 % for MC and 325 % for FM, P< 0·001). Also, when analysing error rate ranges between methods, there were more observations (P< 0·001) with estimation errors higher than 40 % with the MC (56·1 %), than with the FPA (27·5 %) and FM (44·9 %). In conclusion, although differences between estimated and real weight were statistically significant for almost all foods, comparisons between methods showed FPA to be the most accurate tool for estimating food amounts.