Self-host an AI image generation system (based on Stable Diffusion) for your friends

This article introduces advanced usage tips for Stable Diffusion and secure deployment strategies. First, leveraging the PNG Info feature allows you to quickly view and reuse the specific parameters used during image generation; it is recommended to experiment with different combinations of Checkpoints and LoRAs to master stylistic patterns. Second, to address the security risks of exposing local services to the public internet, the article proposes a simplified yet secure UI configuration strategy: disabling VENV to reduce deployment complexity, installing xformers to optimize performance, and modifying startup parameters to hide sensitive directories. Core security measures include using config.json to disable dangerous tabs such as Settings and Train, utilizing ui-config.json to restrict batch generation, fix prompts, and hide the negative prompt input box, thereby preventing resource abuse and the generation of non-compliant content, while simultaneously using user.css to hide irrelevant buttons and simplify the interface. Finally, it is recommended to use FRPC in conjunction with a Caddy reverse proxy to securely expose the service to the public internet, enabling encrypted communication and automatic TLS certificate management, ensuring that friends can conveniently access your AI painting service while protecting privacy and data security.