【Stable Diffusion】ポーズの呪文集プロンプト一覧を紹介!
TLDRThe video provides a comprehensive guide on using 'pose spells' to enhance the variety and quality of image generation through stable diffusion. It explains that entering the correct pose spell is crucial for achieving desired poses, as stable diffusion may not generate them without it. The transcript offers a variety of pose spells, such as 'standing, raise one leg' and 'fold-arms', to help users control the pose in the generated images. It also suggests hiding hands to avoid AI's difficulty with detailed fingers and recommends specific spells for poses like sitting, walking, and lying down. The video emphasizes the importance of aspect ratio in image generation, suggesting vertical for standing or sitting and square or landscape for poses like lying face down. It also introduces the concept of adjusting the spell's influence with a numerical value, recommending a maximum of 1.3 to maintain image quality. The transcript concludes with tips on prioritizing spells in prompts and using 'open pose editor' for more control over pose generation. The video aims to equip viewers with the knowledge to generate ideal poses in their AI-generated images.
Takeaways
- 🔍 Use specific 'pose spells' to increase the variation and likelihood of generating desired poses in Stable Diffusion.
- 🧍 For generating a standing person, use the 'standing' spell, and add 'standing, raise one leg' for a leg up.
- 🤚 Adding 'fold-arms' or 'crossed arms' can help hide hands, which is useful as AI often struggles with detailed fingers.
- 🤲 Including 'arms behind back' or 'hand between legs' can also be effective for hiding hands and improving image quality.
- 🚗 To generate a pose involving a car, use 'hold the steering wheel of a car', which works better with certain aspect ratios.
- 🤸♂️ For dynamic poses like 'reaching out' or 'arms raised in the air', be aware that these can be challenging to generate perfectly.
- 👶 For a child sitting on a chair, use 'sit' or 'sitting', and consider vertical image sizes for balance.
- 🏃 Adding 'walking', 'running', or 'jumping' to the prompt can generate movement poses, though success may require multiple attempts.
- 🛌 To depict lying down or sleeping poses, use 'lie down' or 'sleep', and adjust the aspect ratio for the best results.
- 🧘 For seated poses like 'wariza', 'seiza', or 'lotus position', experiment with different aspect ratios to achieve the desired outcome.
- 📐 If a pose isn't generating as desired, adjust the aspect ratio or use a numerical value after the spell (e.g., run1.3) to increase its influence, with 1.3 being the recommended maximum to avoid quality deterioration.
Q & A
「ポーズの呪文」とは何ですか?
-「ポーズの呪文」とは、Stable DiffusionなどのAI画像生成モデルに特定のポーズを生成させるために入力する指示やキーワードのことです。この呪文によって、望むポーズの画像を効率よく生成することが可能になります。
手を隠すポーズの呪文にはどのようなものがありますか?
-手を隠すポーズには「fold-arms」(腕を交差させる)、「arms behind back」(手を背中で組む)、または「hand between legs」(脚の間に手を置く)などがあります。これらの呪文は手の描写が難しい場合に有効です。
画像生成において、どのようにして手のポーズをより自然に見せることができますか?
-手のポーズを自然に見せるためには、「own hands together」(手を合わせる)などの呪文を利用し、さらに画像のアスペクト比や数値を調整することが推奨されます。これにより、指の不自然な動きや位置を避けることができます。
ポーズを成功させるためのアスペクト比は何ですか?
-ポーズによって最適なアスペクト比が異なります。たとえば、立っているポーズは縦長の比率が生成しやすく、地面に座っているか横になっているポーズは四角形または横長の比率が適しています。
画像生成で高い確率でポーズが生成されるにはどうしたら良いですか?
-高い確率でポーズを生成するには、呪文の数値を調整する(例えば「run1.3」)や、呪文をプロンプトの始めに置くことが有効です。これにより、指定したポーズが優先して生成されやすくなります。
一脚を上げた立ち姿を生成するための呪文は何ですか?
-一脚を上げた立ち姿を生成する場合、「standing, raise one leg」という呪文を使用します。これにより、その特定のポーズを生成する確率が上がります。
モデルが手を腰に置くポーズを生成するための呪文は何ですか?
-手を腰に置くポーズを生成するためには、「put hands on hip」や「one hand on hip」のような呪文を使います。これにより、モデルが腰に手を置くポーズを生成する確率が高まります。
車の運転姿勢を生成するための呪文は何ですか?
-車を運転している姿勢を生成する場合は、「hold the steering wheel of a car」と入力します。また、ポートレートではなく、四角形または横長の画像を使用すると生成が容易です。
Outlines
🎭 Introduction to Pose Spells for Image Generation
The first paragraph introduces the concept of 'pose spells' which are specific prompts used to guide the generation of poses in AI image creation. It emphasizes the importance of using the correct spell to increase the likelihood of achieving the desired pose, such as standing, sitting, or lying down. The paragraph provides various examples of pose spells, like 'standing, raise one leg' or 'fold-arms', and explains how they can be combined for more complex poses. It also discusses the challenge of generating images with hands in certain positions and offers strategies to overcome this, such as hiding hands or using specific spells like 'arms raised in the air'. The paragraph concludes with tips on generating poses for specific scenarios, like holding a steering wheel or jumping, and the importance of aspect ratio in achieving the desired outcome.
🖼️ Tips for Generating the Ideal Pose
The second paragraph delves into strategies for successfully generating the ideal pose when using AI image creation tools. It suggests changing the aspect ratio of the image to vertical, square, or landscape to influence the type of pose that is generated, such as standing or sitting. The paragraph also recommends entering a numerical value after the pose spell, up to a maximum of 1.3, to enhance the spell's influence on the image generation. It advises placing the most important pose spell at the beginning of the prompt for higher priority in generation. The paragraph discusses the unpredictability of pose generation and the need for multiple attempts to achieve the desired result. It also mentions the use of 'open pose' or 'open pose editor' for more control over the pose and provides a video link for further guidance.
🔄 Summary of Techniques for Pose Generation
The third paragraph summarizes the key points and techniques for generating poses in AI image creation. It reiterates the effectiveness of changing the aspect ratio and adjusting the spell's influence with a modifier value for better results. The paragraph emphasizes the importance of prioritizing the desired pose spell by placing it at the beginning of the prompt. It encourages viewers to refer to the provided spells and experiment with generating their ideal poses. The paragraph concludes with a call to action, inviting viewers to subscribe to the channel for more informative content on AI-generated images.
Mindmap
Keywords
💡Stable Diffusion
💡Pose Spells
💡Image Generation
💡Aspect Ratio
💡Hands Positioning
💡Numerical Value Adjustment
💡Priority in Prompts
💡Open Pose Editor
💡Walking and Running
💡Lying and Sleeping Positions
💡Crossed Legs and Lotus Position
Highlights
Stable diffusion often requires the correct pose spell to generate certain poses.
Adding specific pose spells can increase the variation and probability of desired poses.
Combining two spells can generate complex poses, such as standing with one leg raised.
Folding arms or hiding hands can improve image generation quality due to AI's difficulty with fingers.
Adding 'arms behind back' or 'hand between legs' can create natural-looking poses.
Poses like 'put hands on hip' or 'one hand on hip' can be generated with specific spell additions.
Spells like 'hold the steering wheel of a car' can generate poses related to specific actions.
Using 'reaching out' can generate poses with arms extended, though it can be challenging.
Artistic poses like 'arms raised in the air' are often used and can be generated with the right spell.
Aspect ratio adjustments can significantly impact the generation of standing or sitting poses.
Numerical values after a spell can adjust its influence, with a recommended maximum of 1.3.
Placing the pose spell at the beginning of the prompt can increase the likelihood of generation.
The probability of generating a specific pose can vary, sometimes requiring multiple attempts.
Using 'open pose' or 'open pose editor' can help generate poses not covered in the video.
Changing the aspect ratio to vertical, square, or landscape can help achieve different poses.
Landscape and square ratios are better for poses like lying on your side or stomach.
Vertical poses are easier to generate for lying on the back, while horizontal poses are better for side sleeping.
The video provides a comprehensive collection of pose spells for generating a wide range of poses in stable diffusion.