The prediction of physicochemical attributes in foods using visual indicators represents a valuable strategy for non-invasive quality monitoring. In this study, a model was developed to estimate the pH and total titratable acidity (TTA) of a powdered condiment made with Comadia redtenbacheri (chinicuil), based on visual information obtained through computer vision. Images were captured under controlled conditions using a high-resolution 48 MP camera positioned 40 cm above the sample at a fixed 90° angle, inside a standardized lighting chamber equipped with daylight-balanced neutral white LED illumination (5600 K) to prevent shadows and reflections. Subsequently, the color components L*, a*, and b* were extracted using OpenCV. This visual data was integrated with experimental measurements of pH and TTA obtained during storage at different temperatures and times, forming a single structured dataset. Multiple linear regression models (OLS) were fitted to evaluate the relationship between visual descriptors and physicochemical properties. The model for pH showed a moderate fit (adjusted R² = 0.403), with temperature being the only statistically significant predictor. In contrast, the model for TTA achieved a stronger fit (adjusted R² = 0.779), with both time and temperature identified as relevant predictors. None of the color components (L*, a*, b*) had significant individual effects, although they contributed to the overall model structure. These results suggest that the product’s acidity is more sensitive to storage conditions than pH. In conclusion, the study demonstrates the feasibility of using computer vision as a complementary tool to estimate physicochemical properties throughout the product’s shelf life, with potential for integration into automated monitoring systems for dehydrated foods.