Stable Diffusionで高品質な美少女ができるプロンプト呪文集をまとめて紹介

AIジェネ【AIイラスト生成の情報発信】
2 Sept 202313:15

TLDRThe video provides a comprehensive guide on generating high-quality images of beautiful girls using stable diffusion. It introduces various 'spells' or prompts that can be used to enhance the quality of the generated images. The video emphasizes the importance of using a combination of 'masterpiece', 'best quality', and 'detailed' spells to achieve the best results. It also suggests using 'EasyNegative' in the negative prompt to further improve the quality. The video advises against using too many spells, as it may lead to a decrease in image quality. It recommends setting up to three overall quality spells and using specific part-focused spells to correct any issues. The video also covers the use of 'vae' for enhancing image quality and adjusting the sampling method and image size settings for better results. It concludes by recommending viewers to refer to the prompt content from the Civitai site for additional guidance.

Takeaways

  • 📚 Beginners can start with stable diffusion but need to know the right prompts to generate high-quality images.
  • 🔍 The model 'any lora' from the Civitai site is used in the video for generating images.
  • 🎨 The spell 'masterpiece' greatly improves the overall style, hair, and eyes, making it a must-have for high-quality images.
  • ➕ Combining 'masterpiece' with 'best quality' enhances the image quality more effectively than using either alone.
  • 👗 The 'detailed' spell, when combined with 'masterpiece', significantly increases the quality of clothing details.
  • 👀 'High resolution' and 'hires' spells improve the highlights in the eyes, making them look more lively.
  • 📱 The '4k' and '8k' spells increase image quality, with '8k' generally offering better results.
  • 🎮 The 'playstation5' and 'unreal engine 5' spells generate high-quality images, especially when combined with 'masterpiece'.
  • 🖼 For a more realistic, live-action look, use the 'photo realistic' spell along with 'illustration'.
  • ⚠️ Be cautious when using multiple spells; adding too many can decrease the overall image quality.
  • 📝 For higher quality, use up to three overall quality spells and include corrective spells for specific parts that may collapse.

Q & A

  • What is the recommended starting point for beginners who want to create high-quality images using stable diffusion?

    -Beginners can start with the 'masterpiece' spell, which greatly improves the overall style, hair, eyes, and other aspects of the image.

  • What is the difference between using 'master piece' and 'best quality' in the prompt?

    -The 'master piece' spell is considered superior in quality, but combining it with 'best quality' can further enhance the image quality more effectively than using 'master piece' alone.

  • How does the 'detailed' spell affect the quality of the generated images?

    -The 'detailed' spell, when used in combination with 'master piece', can significantly increase the quality of the image, particularly in terms of the details on the clothes.

  • What is the 'EasyNegative' spell and how does it help in generating high-quality images?

    -The 'EasyNegative' spell is a highly recommended negative prompt that can improve the overall quality of the generated image by omitting low-quality elements.

  • What is the effect of using 'high resolution' or 'hires' in the prompt?

    -The 'high resolution' or 'hires' spells can enhance the quality of the image by adding highlights to the eyes and other parts, making them look more lively and detailed.

  • How does the 'photo realistic' spell change the style of the generated images?

    -The 'photo realistic' spell is recommended for those who want to create images that resemble real-life photography rather than anime or cartoon styles, resulting in a more realistic painting style.

  • What is the recommended approach when encountering issues with specific parts of the image, such as broken hands?

    -To fix specific issues, use corrective spells in the negative prompt that target the problematic part, such as 'bad hands', 'bad fingers', or 'missing fingers'.

  • What is the 'vae' and how can it be used to improve the quality of generated images?

    -The 'vae' is a tool that can be used to brighten the overall image and correct the highlights of the eyes and other parts, thus improving the overall quality of the generated image.

  • What is the recommended image size setting for generating high-quality images, especially when including the whole body?

    -Setting one side of the image to 1024 or more pixels and using 'hires.fix' can dramatically improve the quality and prevent issues like blurred images or collapsed eyes.

  • What is the 'unreal engine 5' spell and how does it affect the quality of the generated images?

    -The 'unreal engine 5' spell, when used in the prompt, can generate images with surprisingly high quality, often resembling 3D game characters.

  • What is the recommended strategy for combining multiple quality-enhancing spells in the prompt?

    -It is suggested to use up to three spells that affect overall quality, such as 'masterpiece, best quality', and then add specific part-targeting spells if needed to address any quality issues.

  • How can one ensure they are using the most effective spells for generating high-quality images?

    -Referencing the prompt content from the Civitai site that distributes the model can provide insights on how to generate quality images and help in determining the most effective spells to use.

Outlines

00:00

🎨 Enhancing Image Quality with Specific Spells

This paragraph discusses various 'spells' or prompt entries that can be used to enhance the quality of generated images, particularly when creating high-quality images of beautiful girls using stable diffusion models like 'any lora'. It emphasizes the importance of using specific prompts such as 'masterpiece', 'best quality', 'detailed', and 'high resolution' to improve different aspects like overall style, hair, eyes, and clothing details. The video also covers the impact of using half-width characters and the potential benefits of combining certain spells for better results. It advises beginners to experiment with different prompts and spells to achieve the desired image quality.

05:01

📸 Photorealistic and 3D Character Generation

The second paragraph explores advanced techniques for generating images with high photorealism and 3D character qualities. It introduces the use of 'unreal engine 5' and 'photo realistic' spells for creating images resembling 3D game characters and live-action styles. The video recommends combining these with 'masterpiece' for enhanced quality. It also discusses the challenge of selecting the right spells and the strategy of using 'detailed' or 'ultra detailed' within the prompt for better results. The paragraph addresses common issues like broken eyes or elongated body parts and suggests using corrective spells in the negative prompt to fix such issues. It highlights 'EasyNegative' as a highly recommended spell to improve overall image quality significantly.

10:05

🖼️ Advanced Techniques for High-Quality Image Generation

The final paragraph offers a summary of the techniques and spells introduced in the video for generating high-quality images. It advises using a combination of 'masterpiece', 'best quality', and 'detailed' spells in the prompt while cautioning against using too many spells, which may reduce quality. The video suggests a maximum of three overall quality spells and emphasizes the effectiveness of negative prompts to exclude low-quality elements. It also recommends using 'EasyNegative' to further enhance image quality. Additional tips include using VAE for improved brightness and highlight correction, selecting appropriate samplers like those ending with 'karras', adjusting image size for better detail, and referring to the model distributor's prompt content for guidance. The video concludes with a reminder to subscribe for more informative content on generation AI.

Mindmap

Keywords

💡Stable Diffusion

Stable Diffusion is a term used in the context of AI-generated images, referring to a method or model that creates stable and high-quality outputs. In the video, it is the starting point for beginners looking to generate high-quality images of beautiful girls. It's mentioned as a foundational tool that, when combined with the right 'spells' or prompts, can produce desired results.

💡Prompt

A prompt in this context is a set of instructions or a phrase that guides the AI in generating an image. It's a core concept in the video as it determines the output's style and quality. The script discusses various prompts that can be used to enhance the image generation process, such as 'masterpiece' and 'best quality'.

💡Spell

In the video, 'spell' is a metaphor for a specific type of prompt or a set of keywords that, when entered into the AI image generation system, influence the quality and style of the generated image. Spells like 'masterpiece', 'best quality', and 'detailed' are used to improve the overall aesthetic of the output.

💡Quality

Quality, in the context of the video, refers to the visual and stylistic excellence of the AI-generated images. It's the primary goal when using spells and prompts, with the video providing various methods to enhance this quality, such as combining certain spells or using specific models like 'any lora'.

💡Resolution

Resolution is a term that denotes the level of detail an image can display, typically measured in pixels. In the video, higher resolution spells like 'high resolution' and '8k' are recommended to improve the sharpness and clarity of the generated images, making them more lifelike and detailed.

💡Negative Prompt

A negative prompt is a set of instructions used to guide the AI away from generating certain undesired features or qualities in the image. The video suggests using negative prompts to correct issues like 'bad hands' or 'long neck' by specifying what to avoid, thus improving the final image's quality.

💡Model

In the context of AI image generation, a model refers to a specific AI system or algorithm that processes the prompts to create images. The video mentions 'any lora' as a downloadable model from the Civitai site, which is used to generate high-quality images when combined with the appropriate spells.

💡Civitai

Civitai is a platform mentioned in the video where users can download AI models like 'any lora' for image generation. It serves as a resource for those looking to enhance their AI-generated content, providing models that can be used with various spells to achieve high-quality results.

💡Sampling Method

The sampling method refers to the technique used by the AI to generate the final image from the data provided by the prompts. The video suggests that changing the default 'sampling method' to one ending with 'karras' can lead to higher quality images, indicating the importance of this setting in the image generation process.

💡vae

VAE, or Variational Autoencoder, is a type of AI model used for generating images. In the video, using a VAE is suggested as a way to improve the overall quality of the generated images by brightening the whole image and correcting the highlights of the eyes and other parts.

💡Image Size

Image size is a crucial factor in determining the clarity and detail of the generated image. The video emphasizes the importance of setting a larger image size, such as 1024 pixels or more, and using 'hires.fix' to prevent common issues like blurry images or distorted features.

Highlights

Beginners can start with stable diffusion to generate high-quality images of beautiful girls.

Using the 'masterpiece' spell can greatly improve the overall quality of the generated image.

Combining 'masterpiece' with 'best quality' enhances the quality more effectively than using 'masterpiece' alone.

The 'detailed' spell, when combined with 'masterpiece', significantly increases image quality, especially in clothing details.

The 'all detailed' spell alters the details of the clothes compared to using only the 'masterpiece' spell.

The 'ultra detailed' spell produces quality similar to 'detailed', offering an alternative choice for image generation.

The 'extremely detailed' spell generates images with more flashy clothing, suitable for those focusing on clothing decoration.

The 'extremely detailed cg unity 8k wallpaper' spell is less detailed but provides a good overall image quality.

The 'intricate details' spell is recommended for creating illustrations with realistic hair.

Adding 'high resolution' to the prompt results in better highlights in the eyes, making them appear more lively.

The '4k' spell combined with 'masterpiece' improves image quality, while '8k' tends to produce even higher quality images.

The 'playstation5' spell generates very high-quality images, especially when combined with 'masterpiece'.

Entering 'unreal engine 5' in the prompt can generate surprisingly high-quality images resembling 3D game characters.

The 'photo realistic' spell is ideal for those who want to create a more live-action look for their beautiful girls.

It is recommended to use up to 3 spells affecting overall quality, such as 'masterpiece, best quality', for better results.

Using corrective spells in the negative prompt can fix issues with specific body parts, like hands or fingers.

The 'EasyNegative' spell in the negative prompt can significantly improve the overall quality of the generated image.

Using VAE can enhance the brightness and correct highlights, further improving the image quality.

Changing the sampling method to one ending with 'karras' can help generate higher quality images.

Increasing the image size to 1024 or more and using 'hires.fix' can dramatically improve image quality, despite longer generation time.

Referencing the prompt content from the Civitai site can provide valuable guidance on generating high-quality images.

In summary, adding 'master piece, best quality, detailed' to the prompt is recommended for generating high-quality images.

Create Stunning AI Art Instantly

No login. No cost. Just incredible AI images !

Create Stunning AI Art InstantlyGenerate Now