A shiny silver evening gown

My model, all public welfare, free of charge, prohibited for commercial use, please conscientiously abide by local laws and regulations when using. If you like it, remember to give it a thumbs up and return some points. Thank you for your support

Additional Notes: If you find that my Lora model, everyone else and the author are using it well, you added triggering words, and the usage is correct, but still has no effect at all, it is likely because I am modifying the usage instructions for Lora, and if it's under review, it won't take effect. Each modification will be reviewed in about half an hour to an hour, maybe faster, like 10-20 minutes, and it will return to normal shortly after.

Reference Weight: 0.7       Card drawing required     Adjustment range 0.6-0.8    

Core Trigger Words:choker, necklace, bare shoulders, large breasts, cleavage, sleeveless, white dress, high heels,

Alternative Words:wet, see-through,Optional Words:may or may not be added

Recommended Models:

Realistic: majicMIX realistic Orange Writing Realistic_v7（My favorite universal realistic alchemy model, other models look fancy, after alchemy, compatibility is poor, many bugs, not universal. Of course, this is my personal preference, no intent to offend others.）

Anime: AstrAnime 2.5D Anime（Excellent transparent wet effect presentation）, Extreme Comic Mix (Vivid Colors), Bold 2.5D by White City Lord (Similar to thick-lined comics), YesMix for Girls v1.6 (original), TMND-Mix (With VAE, fresh and bright colors), GuoFeng3 Chinese Style 3 V3.4, sexyshojokFantaxiMIX (A model between 2.5D and 3D, strong in displaying wet effects) etc., all have good performance. Some models may require VAE addition, check the author's page for specific instructions.

Realistic, Anime, 2.5D all acceptable

Note: For this model, whether it is a face swap, seated, half-body, or full-body image, 512x1024 resolution is recommended, 768x1024 is not ideal. However, 768x1024 can also be tried, it's just a bit luck-based for the output, but for face swaps, 768x1024 would have an advantage.

If you use high-definition restoration, the image texture and output rate will be significantly improved, recommended upscaling algorithm R-ESRGAN_4x+, restoration weight 0.33,For anime-style artwork, choose the corresponding upscaling algorithm R-ESRGAN_4x+ Anime6B, restoration weight 0.33. Experienced users may consider using 0.35 or 0.75, but results are not guaranteed. Based on current tests, I recommend the restoration weight of 0.33.

For many beginners, let me say this: Examples are for reference only, but without adding the triggering words I provided, the image will not be generated. Also, it is recommended to use concise quality description terms. AI, like humans, gets more distracted with more description terms, making it harder to focus and reducing the expressive power. Just like I do, two or three key quality description terms are sufficient. And don't add a bunch of lora unless compatible, most lora are not compatible with each other, diluting attention and disrupting weight distribution, let alone the incompatibility causing failure. Also, avoid choosing so-called impressive base models that come with various filters and special visual effects, which may cause Lora to be incompatible, except for experienced users. For beginners, the first priority when selecting a large or base model is compatibility; if it's not compatible, it's all in vain. The more compatibility, the more lora options you have. Unless you're using that model as a base, specifically training lora for it. When generating images with lora, the top priority is still the models recommended by the author after testing, or the base models used for training lora. Experienced users need not pay attention, this advice is for newcomers stepping into AI art.

Example 1: (Realistic)

Front View:

1 girl, solo, perfect face, eye contact, sitting,

choker, necklace, bare shoulders, large breasts, cleavage, sleeveless, white dress, high heels,

masterpiece, best quality, highly detailed,

Negative: various faces, explicit content, multiple characters, poor quality elements, low resolution, etc.

Parameters: Steps: 30, Sampler: DDIM, CFG scale: 7.0, Seed: 2898232163, Size: 512x1024, Model hash: 7c819b6d13, Model: majicMIX realistic Orange Writing Realistic_v7.safetensors, Denoising strength: 0.33, Clip skip: 2, RNG: CPU, Lora 1: xxxxx, Lora Hash 1: 6839e19c29, Lora Weight 1: 0.7, vae_name: automatic, Hires resize: 1024x2048, Hires steps: 30, Hires upscaler: R-ESRGAN_4x+

Example 2: (Anime)

Front View:

1 girl, solo, perfect face, eye contact, sitting,

choker, necklace, bare shoulders, large breasts, cleavage, sleeveless, white dress, high heels,

masterpiece, best quality, highly detailed,

Negative: various faces, explicit content, multiple characters, poor quality elements, low resolution, etc.

Parameters: Steps: 30, Sampler: Euler a, CFG scale: 7.0, Seed: 699774074, Size: 512x1024, Model hash: 5b53949fc6, Model: AstrAnime 2.5D Anime_V6.safetensors, Denoising strength: 0.33, Clip skip: 2, ENSD: 31337, RNG: CPU, Lora 1: xxxxx, Lora Hash 1: 6839e19c29, Lora Weight 1: 0.7, vae_name: automatic, Hires resize: 1024x2048, Hires steps: 30, Hires upscaler: R-ESRGAN_4x+ Anime6B