The section outlines Training-Free Guidance (TFG), an innovative framework for conditional content generation using generative models, particularly diffusion-based ones. TFG enables guiding the generative process without the need to retrain the model, leveraging pre-existing "predictors" to steer generation toward specific characteristics, enhancing efficiency and flexibility. This approach unifies various existing methods into a single configurable space, which can be optimized through hyperparameters. TFG has demonstrated superior performance across multiple applications, including images, audio, and molecules, outperforming traditional methods in terms of accuracy and efficiency, while unlocking new possibilities across different fields.