【総集編】これ1本でStable Diffusionがわかる！ PCの選び方～インストール～拡張機能の使い方まで初心向けに2023年を総まとめして徹底解説

This paragraph introduces a video script focusing on the evolution of AI image generation in 2023, particularly highlighting the 'Stable Diffusion' model. The script discusses the creation of cute illustrations using AI, with a year-end special that reviews the advancements in image generation AI throughout the year. It mentions the installation of Stable Diffusion, choosing the right PC, and the use of extensions like Lola and ControlNet. The script is aimed at beginners in AI image generation, offering a detailed guide from installation to creating illustrations. It also touches on the channel's video production changes and the evolution of the character 'Sefirin-chan' over time.

The second paragraph delves into the specifics of selecting a computer for AI image generation, emphasizing the importance of a powerful GPU. It discusses the decision between investing in a gaming PC with a strong GPU or purchasing an external GPU. The script recommends a gaming PC with an RTX 3060 for beginners, noting its cost and adequacy for entry-level AI image creation. It also touches on the evolution of the channel's approach to creating videos and the changes in the character 'Sefirin-chan', suggesting that viewers watch the video to the end to understand the progression of 'Stable Diffusion' in 2023.

Paragraph three provides an in-depth look at choosing a graphic card for AI art generation. It explains the significance of a graphic card, comparing it to an artist's canvas and brushes. The discussion includes the importance of VRAM size for handling large images and multiple image generation. The script guides viewers on how to determine the necessary VRAM size by demonstrating the process with different image sizes and resolutions. It also covers the variety of graphic cards available, such as the GTX 16 series, RTX 30 series, and RTX 40 series, with a recommendation for RTX 3060 with 12GB VRAM as a suitable choice for beginners. The paragraph cautions against purchasing the wrong type of GPU, specifically advising against AMD's Radeon series for 'Stable Diffusion'.

The fourth paragraph discusses the selection of CPU, memory, and storage for a PC used in AI image generation. It explains the role of the CPU as the 'brain' of the computer and its lesser impact on image generation compared to other tasks like gaming or video editing. The script advises on choosing between Intel and AMD CPUs, focusing on performance and cost-effectiveness. It also covers the importance of memory (RAM) for ongoing processes and storage for saving data, recommending 16GB of RAM and a combination of SSD and HDD for storage. The paragraph provides a comprehensive guide on balancing these components for optimal performance in AI image generation.

Paragraph five offers advice for purchasing a PC for AI image generation, recommending BTO (Build-to-Order) shops and gaming PCs. It explains the benefits of BTO, such as customization options, and the importance of selecting a gaming PC due to its graphic card capabilities. The script provides a checklist of recommended specifications, including an RTX 3060 GPU, a CPU from the Core i5 or Ryzen 5 class, 16GB of RAM, and a 500GB SSD. It also touches on the considerations for self-building a PC, concluding that BTO is a more accessible option for beginners.

The sixth paragraph outlines the process of installing 'Stable Diffusion' and preparing for AI art creation. It covers the prerequisites, such as a computer with a compatible graphic card, and the software requirements including Python, Git, and the 'Stable Diffusion' web UI. The script provides a step-by-step guide to installing these components, emphasizing the importance of checking the installation of Python and Git before proceeding. It also mentions the need for a graphic card with at least 10GB of VRAM, specifically recommending the RTX 3060, and provides a link to the 'Stable Diffusion' web UI in the video description for further guidance.

Paragraph seven discusses the creation of AI-generated cosplay images and the importance of 'Lora' in enhancing the process. It mentions a video from April 11, 2023, which detailed the use of 'Lora' for creating AI cosplay. The script also describes the process of generating images using 'Stable Diffusion,' including the use of prompts to guide the AI in creating specific illustrations. It touches on the challenges of generating images of a character named 'Shiranui Ma' and the iterative process of refining the prompts and learning to achieve the desired result.

The eighth paragraph explores the impact of hair color on AI-generated images and the concept of additional learning or 'Lora' to improve the generation of specific characters. It describes an experiment where different hair colors are specified to see how they affect the overall image. The script also discusses the idea of additional learning with 'Lora' to better generate images of characters not commonly found in the AI's training data. The paragraph concludes with the successful creation of an image resembling the character 'Shiranui Mai' through the use of 'Lora' and additional learning.

Paragraph nine delves into the use of ControlNet for pose control in AI-generated illustrations. It mentions a video from April 29, 2023, which discussed the capabilities of ControlNet in creating poses for illustrations. The script describes the process of creating an AI illustration of the character 'Shirei' and the challenges in posing. It explores various methods within ControlNet for pose specification, such as 'Canny' and 'Line Art,' and their effectiveness in creating illustrations that match the desired pose. The paragraph concludes with the successful generation of a pose for 'Shirei' using ControlNet.

The tenth paragraph discusses the creation of 'Star of Love' illustrations using a feature called 'Scribble' in AI art generation. It explains the challenges in creating a specific pose using AI and how 'Scribble' can be used to overcome these challenges. The script describes the process of creating a rough sketch and then using 'Scribble' to generate a detailed illustration based on the sketch. It also touches on the use of ControlNet and the 'Invert from White Black Line' setting to improve the illustration's accuracy. The paragraph concludes with the successful creation of an illustration that matches the desired pose and style.

Paragraph eleven focuses on creating AI cosplay images of fighting game characters, specifically from Street Fighter and KOF. It discusses the process of generating individual character images using prompts and the use of 'Lora' for additional learning to enhance the accuracy of character features. The script also covers the selection of models for generating realistic or illustrative images and the importance of understanding the characters' physiques, such as height and weight, for creating a cohesive group illustration. The paragraph concludes with the preparation of character images for a group illustration that features all the characters in a fighting stance.

The twelfth paragraph describes the process of composing a fantasy illustration featuring female fighters from various games, akin to a group shot of characters from different universes. It details the steps taken to create a cohesive image, including generating individual character images, adjusting poses, and finally compositing them into a single scene. The script also mentions the use of 'ControlNet' and 'Open Pose' to ensure the characters' poses are consistent and the overall image is aesthetically pleasing. The paragraph concludes with the successful creation of a composite image that brings together the distinct characters into one fantastical scene.

Paragraph thirteen discusses the creation of a group cosplay illustration using AI art tools, focusing on the characters Chun-Li, Cammy, and others from fighting games. It covers the process of generating individual character images, selecting appropriate poses, and then compositing these into a group illustration. The script also touches on the use of 'ControlNet' for pose control and finalizing the composite.