NSFW Image-to-Image: Reference-Based Generation for Explicit Photoshoots

In this lesson, we examine the Image-to-Image (generation by reference) tool. Its main purpose is to transfer the pose, style, mood, or composition from one image to your character while preserving the necessary body details and appearance.

This tool is especially useful when:

• We like a specific pose or style in someone else's image
• We want to recreate the scene but with our model
• We need to quickly get ideas and variations without complex prompts

The Essence of Image-to-Image

The tool's logic:

1. We upload a reference image — a liked pose, successful shooting style, specific composition
2. The network generates a similar scene but with our character, with their face and body

We're talking not about copying but visual inspiration.

Models in Image-to-Image

SDXL (Safe Work / Not Safe for Work)

• Can upload a very large number of references
• For each reference, the model generates four images
• Can use LoRA for the body (increase breasts, change proportions)

Limitation: SDXL is not designed to preserve consistent character body.

Use SDXL for experiments, finding poses and ideas, as a source of references for further work.

General Safe Work / Not Safe for Work

This is the main working tool.

• Upload references for poses and style
• Additionally upload a photo of our model (preferably with full body)
• Can preserve pose and change or keep background

The General model produces more realistic and lively photos, especially if references are real photographs.

Practical Recommendations

1. Main model — General Safe Work / Not Safe for Work
2. References — any liked photos (other models, scenes, poses)
3. Your model's photo — nude, with those details that are important to preserve
4. Use SDXL for experiments and idea generation
5. After generation — refine the result with other tools

What to Do with Results

The generated images can be:

• Edited through Editor / Combiner
• Improved with upscaling
• Sent to FaceSwap if the face isn't perfect
• Run through Carousel to get variations

Image-to-Image acts as a bridge between the idea (reference) and final content.