【AI】Leonardo AI界面与功能详细教程 2024 | 零基础入门

The video begins with an introduction to the Leonardo website, where users are guided to register an account or log in using a Google account. The main interface of Leonardo is explored, highlighting its various sections such as Community Feed, which displays images generated by users but with a temporarily disabled search function. The video explains the different search criteria available for finding specific types of images. It also details the process of generating new images based on prompts and parameters, and how to use the Motion and Image-to-Image features. Personal Feed is introduced as a space to save images generated by the user's account. The Training & Datasets section is briefly mentioned as the place to train models, and the different categories of models are outlined: Finetuned Models, Platform Models, Community Models, and User's Models. User Tools are also introduced, including Image Generation, Realtime Canvas, Canvas Editor, Universal Upscaler, Motion, and Texture Generation. The focus then shifts to Image Generation, demonstrating how to remix an existing image, input prompts, and choose models and styles to create new images.

The video delves deeper into the Image Generation feature, showcasing how to generate an image by remixing an existing one and customizing the prompt. It explains how to improve the prompt for better results and introduces the concept of choosing different models, each with varying token requirements. The addition of styles to an image is demonstrated, highlighting the ability to select multiple styles and how these styles can change depending on the model chosen. LoRA models are mentioned as a means of achieving these styles. Negative prompts are also discussed, explaining how to exclude unwanted elements from the generated images. The video then covers the customization of elements within an image, such as turning styles on or off and adjusting their weights to create a blend of styles. The history of generated images and their parameters are shown, along with options to regenerate, view information, copy seed values, and make images public. The video also touches on the number of images to generate at once, the Photoreal, Alchemy, and Prompt Magic enhancement technologies, and their increasing levels of strength and detail. A demonstration of Prompt Magic's effect on image quality is provided, followed by an exploration of Alchemy and Photoreal, showcasing their capabilities in generating high-quality images.

The video concludes with an overview of advanced settings and additional features in image generation. Transparency is introduced as a new feature for creating images with transparent backgrounds using the LayerDiffusion plugin. Public Images is explained as an option to either keep images private or share them to the Community Feed. Customizing image dimensions is demonstrated, with options to input specific values, use sliders, or select from common presets. Advanced Controls allow for specifying the aspect ratio and adjusting the Guidance Scale, which is akin to the CFG value in stable diffusion, affecting the precision or creativity of the generated image. The Tiling feature for creating repetitive textures or backgrounds is briefly mentioned. The video also discusses the Image Guidance function, which has been moved to the Image Guidance tab and will be covered in a future video. Recommended sizes for models are explained, ensuring that the image dimensions are optimal for the chosen model. The fixed seed option is demonstrated, showing how it can generate identical images given the same parameters and model. The Scheduler, akin to a sampling algorithm, is introduced, and the video ends with a teaser for upcoming content on image-to-image functionality and prompt generation.