In my spare time, I deleted the model and retrained it, fine-tuning the weights. It’s highly stable, with reclassified natural language and tag vocabularies to eliminate weight contamination.

Combining photo backgrounds with anime often results in junk plastic content. Therefore, I used natural language to enhance style stability and tagged sensitive content, reducing training volume to rebuild anatomical accuracy in limbs.

Compared to the previous version, this one is more mature.

Description:

i5_Water is designed for training LORA with photo backgrounds and sensitive content, focusing on details between backgrounds and water.

Recommended Tools:

ComfyUI: Local video generation. It may have bugs but is free and ideal for high-quality image/video output. Save money and invest in things that bring returns.

Stable Diffusion A1111: Local image generation. It doesn’t support Wan Video but is the best branch tool. No need to spend extra time on detailed edits—just use prompts.

Summary:

Free is key because you need protein and vitamins to maintain your body. Save money for that. Don’t fall for fancy promises like diet foods or premium subscriptions making you slim, or the “best AI app” making you rich. Without something as powerful as GPT, I wouldn’t subscribe to anything randomly.

Note:

Imagine free resources as precious and beautiful as abundant water. Gold may be valuable, but without human services, it can never meet your physiological needs.