In order to improve cassava's palatability and reduce its toxicity, this root is fermented and applied in foods, such as sour cassava starch used to prepare cheese bread and biscuits. This fermentation occurs spontaneously with lactic acid bacteria (LAB) and yeasts. However, it remains an empirical process, with long duration and lack of product quality homogeneity. This work aims to use starter cultures in a pilot-scale fermentation process for the production of sour cassava starch. After differentiation of strains, Lactobacillus plantarum Lp3, which exhibited great total titratable acidity (TTA) (5.01 ± 0.05%) and antagonistic activity against Bacillus cereus, Escherichia coli and Salmonella Typhimurium, together with Lactobacillus brevis Lb9 (with lesser TTA values: 2.71 ± 0.10%, but amylolytic activity: 2.75 ± 0.61 mm) were tested as single and co-cultures with Saccharomyces cerevisiae UFMG-A1007. LAB and yeasts were inoculated at counts of 8 and 7 log 10 CFU/g, respectively, and they remained until the 28th day only in co-culture, highlighting the importance of the yeast for the LAB viability. Although single cultures lead to higher acidity during fermentation, the final product acidity obtained with single cultures did not differ from the acidity obtained with L. plantarum Lp3 in association with S. cerevisiae UFMG-A1007. Therefore, this co-culture exhibited higher potential to be tested as a starter culture in industrial-scale fermentation studies because both microorganisms were in high counts until the end of fermentation and contributed to a final product safe for human consumption, with satisfactory acidity, expansion capacity, and physicochemical properties.