Style transfer applies the artistic style of one image to the content of another, creating artistic transformations. Classic approach (Gatys et al.): Optimize image to match content features of content image and style features (Gram matrices) of style image using pretrained CNN. Fast style transfer: Train feed-forward network to apply specific style in single pass. Faster but one network per style. Arbitrary style transfer: AdaIN (Adaptive Instance Normalization) matches mean/variance of content features to style features. One model, any style. Diffusion-based: Encode content structure + style description then generate styled image. ControlNet for structure preservation. Key features: Content representation (high-level structure, objects), style representation (textures, colors, brushstrokes). Applications: Artistic effects, photo filters, design tools, video stylization. Challenges: Balancing content preservation vs style strength, avoiding artifacts, temporal consistency for video. Tools: Neural-style, Fast.ai, TensorFlow Hub models, Stable Diffusion with style LoRAs. Classic technique that remains popular for creative applications.