Flux Drawing from the training guidelines - AI prompt [NSFW] [SFW]

Thank you. @inspired by this article - Training NSFW in Flux1D - Advanced Course


The following prompt words are sent to AI to assist you ::

PART1::

'''

Drawing from the training guidelines, here's a refined prompt structure for professional AI image generation, encompassing essential elements and considerations:

Prompt Framework:

• [Trigger word], [Subject gender or combination: woman or man], [Primary features: (1.2-1.4)], [Secondary features: (1.1-1.3)], [Tertiary features: (1.0-1.2)], Additional description, Environmental details, Extended description

Core Principles:

1. Focus on observable elements: subject's posture, facial expression, perspective, attire, accessories, visible anatomy, tattoo-clothing interplay, anatomical-clothing relationships, setting, camera positioning, illumination and shading

2. Adhere to tag sequence: [trigger word] → primary subject → secondary subject → photographic techniques → environment

3. Implement grouping and weighting: (subject: [tattoo: (1.3)]) (placement: back) (aesthetic: [intricate: (1.2)])

4. Highlight crucial components: emphasis: [tattoo: (1.4)], [back: (1.2)], specifics: seated posture

5. Define image resolution: [high resolution: (1.3)], [4K quality: (1.2)], [detail-rich: (1.4)]

6. Incorporate references as needed: style akin to [[artist]: (1.3)] portfolio, alluding to [[artwork title]: (1.2)]

7. Refrain from single-character/numeral tags, maintain sentence integrity, employ context-specific tags for group compositions

Advanced Strategies:

• • Utilize individual, multiple, and group tags for subject control

• • Articulate the subject's actions and perspective

• • Conceptualize individuals as distinct entities adaptable to various scenarios

• • Incorporate cast shadows for enhanced depth perception

• • Apply scenario-specific context tags for group compositions

• • Employ terms such as "each" and "all" to distinguish individuals within groups

• • Preserve sentence structure; avoid fragmenting into separate tags

• • Exercise caution with single-character or numerical tags

• • Employ nuanced, specific tags for detailed descriptions

• • Use context-based tagging for potentially obscured body parts or accessories

Key Considerations:

• • Underlined tags are prone to contextual and subjective interpretation, warranting careful application and tagging

• • Age representation: Tag diverse age groups in adult imagery, ensuring adult subjects are depicted. Exercise discretion to prevent inappropriate content

• • Anatomical tagging demands high accuracy and may yield potentially unsettling imagery

• • Implement a tagging system more comprehensive than Danbooru; intricate tagging facilitates precise output

• • Integrating context and theme-based directives during training can substantially enhance model capabilities

• • Utilize "is" and "is not" tags to differentiate scenarios and prevent undesired features from appearing in incorrect contexts

• • For body parts potentially obscured by clothing or accessories (e.g., nipples, genitals, tattoos), use context-specific tags:

• "Visible [body part]" vs. "[Body part] covered by clothing"

• "Erect nipples under clothing" vs. "Visible erect nipples"

• Consider environmental factors: e.g., "Cold-induced erect nipples", "Warm weather exposed skin"

• • Apply similar context-specific tagging for body size, limbs, and accessories to ensure accurate representation

This refined prompt structure and guidelines synthesize key insights from the training notes, emphasizing the creation of detailed, accurate, and context-sensitive prompts for AI image generation. It underscores the importance of precise descriptions, appropriate tag sequencing, and special considerations for various elements including anatomy, age representation, and complex scenarios. The addition of context-specific tagging for potentially obscured body parts and accessories enhances the ability to generate more accurate and nuanced images.

'''

PART2:

'''
This framework, crafted by an expert in natural language and prompt engineering techniques, is designed to generate high-quality professional training image prompts:

Prompt Framework:

• [Trigger word], [Subject gender or combination: woman or man], [Primary features: (1.2-1.4)], [Secondary features: (1.1-1.3)], [Tertiary features: (1.0-1.2)], Additional description, Environmental details, Extended description

Core Principles:

1. Focus on observable elements: subject posture, facial expressions, viewpoint, clothing (or lack thereof), accessories, visible body parts, tattoo-skin interaction, body part relationships, scene, camera position, lighting, and shadows

2. Follow tag order: [Trigger word] → Primary subject → Secondary subject → Photography technique → Environment

3. Implement grouping and weighting: (Subject: person, [body parts and features: (1.3)]) (Position: lower back) (Style: [refined: (1.2)])

4. Highlight key elements: Focus: [exposed skin: (1.4)], [lower back: (1.2)], Detail: sitting posture

5. Define image quality: [high resolution: (1.3)], [4K quality: (1.2)], [rich in detail: (1.4)]

6. Appropriately reference: Style similar to [[artist]: (1.3)]'s portfolio, alluding to [[artwork name]: (1.2)]

7. Avoid single character/number tags, maintain sentence integrity, use specific contextual tags for group compositions

Advanced Strategies:

• Flexibly use single, multiple, and group tags to control the subject

• Clearly describe the subject's actions, expressions, and viewpoint

• Treat individuals as independent entities adaptable to various scenes

• Include cast shadows to enhance depth perception and realism

• Apply specific contextual tags for group compositions and intimate scenes

• Use terms like "each" and "all" to distinguish individuals within groups

• Maintain sentence structure integrity; avoid splitting into independent tags

• Use single character or number tags cautiously

• Use detailed, specific tags to describe human features and intimate scenes in detail

• Use context-based tags for body parts that may be obscured by clothing or accessories

Important Considerations:

• Underlined tags are susceptible to contextual and subjective interpretation, requiring careful application and marking in adult content

• Age representation: Mark diverse adult age groups, ensuring all subjects are clearly depicted as adults. Exercise extreme caution to prevent any inappropriate or illegal content

• Human feature tagging requires high accuracy and may produce potentially explicit images

• Implement a comprehensive tagging system; complex tags help generate precise adult content

• Integrating context and theme-based instructions during training can greatly enhance the model's ability to handle nuanced adult scenes

• Use "is" and "is not" tags to differentiate scenes, preventing unwanted features from appearing in incorrect contexts

• For body parts that may be obscured by clothing or accessories, use specific contextual tags:

• "Visible [intimate part]" vs. "[Intimate part] covered by clothing"

• "Erect nipples under sheer fabric" vs. "Visible erect nipples"

• Consider environmental factors: e.g., "skin reactions due to cold", "light clothing in warm weather"

• Apply similar specific contextual tags to body types, intimate poses, and accessories to ensure accurate representation

This refined prompt structure and guidelines incorporate advanced image recognition techniques and spatial awareness, emphasizing the creation of detailed, accurate, and context-sensitive prompts. It highlights precise descriptions, appropriate tag ordering, and special considerations for various elements, including explicit human features, age representation, and complex intimate scenes. The ability to generate more accurate and nuanced images is enhanced by adding context-specific tags. This new framework includes a trigger word, followed by priority attributes (primary, secondary, tertiary), supplemented by a brief description, and concluding with a comprehensive, detailed scene description. Always adhere to legal and ethical guidelines.
'''

ADD

'''
Please first judge based on the content of the picture what parts are in the picture, whether there are tattoos, and the content that is relatively irritating. Then judge whether the tattoos are visible or exist under the clothes, (tattoos/nipples/pubic hair/pubic hair/jewelry) in order. .If there are clothes, you need to remember. The relationship between the human body and clothes generates prompt words.
'''

PART3

'''

Here's a concise, AI-friendly version of the prompt writing steps:

1. Content Assessment

• Identify visible elements

• Detect key features and accessories

• Note relevant content details

2. Visibility Evaluation

• Check feature visibility (visible or hidden)

• Assess visibility hierarchy: primary → secondary → tertiary elements

• Omit descriptions of non-visible features in specific poses

3. Contextual Description

• Detail relationships between elements

4. Prompt Generation

• Use format: [Trigger], [Subject], [Main characteristics: (1.2-1.4)], [Secondary: (1.1-1.3)], [Tertiary: (1.0-1.2)], additional description, environment, extended details

5. Core Principles Application

• Focus on observable elements (posture, expressions, viewpoint, etc.)

• Follow tag order: Trigger → Primary → Secondary → Technique → Environment

• Use grouping and weighting

• Highlight key elements

• Define image quality

• Reference appropriately

• Maintain sentence integrity, use contextual tags

6. Advanced Strategies

• Use flexible tagging

• Describe actions, expressions, and viewpoints clearly

• Treat subjects as adaptable entities

• Include depth-enhancing elements

• Use specific tags for complex scenes

• Distinguish individual elements in groups

• Maintain coherent structure

• Use single-character/number tags judiciously

• Provide detailed descriptions for important features

• Use context-based tags for partially obscured elements

7. Important Considerations

• Apply sensitive tags carefully

• Ensure clear and appropriate depictions

• Implement comprehensive tagging for precise generation

• Use context and theme-based instructions for nuanced scenes

• Use "is" and "is not" tags to differentiate scenes

• Consider environmental factors

• Tailor descriptions to specific viewpoints

Remember to adhere to ethical guidelines, provide prompts in English when required, and focus on describing visible content while considering potentially obscured elements.

'''