Image Generation: guides & tutorials

Scraping collects reference images, style samples, and product photos that inform prompts or provide training data for custom models. Apify actors can download and organize image datasets from e-commerce sites, design galleries, or social platforms, building the raw material that fine-tuning and ControlNet workflows need.

Copyright law around AI training data is actively litigated. Images are creative works with full copyright protection. Training on scraped images without a license may constitute infringement depending on jurisdiction and fair use analysis. Getty Images, Shutterstock, and others have sued AI companies over training data. Always involve legal counsel.

Apify collects and stores images; ComfyUI or AUTOMATIC1111 processes them for generation. Python scripts using Pillow preprocess collected images for training. Hugging Face datasets hosts organized collections. The pipeline: scrape → validate → organize → preprocess → train or prompt, each stage with its own quality checks.

Extract image URLs from product pages alongside metadata—SKU, category, color attributes. Download asynchronously with async HTTP clients, validate formats, and organize by category. Apify datasets store metadata while key-value stores or S3 hold binary files. Maintain a manifest mapping SKUs to file paths for downstream processing.