Soil pH is vital in agriculture and environmental science, affecting plant growth, nutrient availability, and soil health. Remote sensing and Machine Learning (ML) accurately predict and map soil pH, guiding effective soil management for optimized yields and environmental conservation. This paper compares ML-based soil pH prediction using multispectral bands and various spectral indices derived from remote sensing data. It uses indices such as vegetation indices, soil indices, temperature, and thermal indices individually and in combination to predict soil pH. Three primary ML models, namely Partial Least Square Regression (PLSR), Support Vector Regression (SVR), and Multilayer Perceptron (MLP), are utilized for predicting soil pH. The findings suggest that spectral indices yield marginally superior predictions compared to spectral bands alone. This finding highlights the significance of using spectral indices, especially vegetation indices, to estimate the soil pH. The incorporation of spectral indices improves the forecast precision because of the subtle data they record about the characteristics and composition of the soil. This research advances remote sensing in soil science by shedding light on the viability and effectiveness of using satellite data for soil pH prediction.