Here's my attempt.

Works well with the following parameters:

• CFG Range: 6-8

• Prompting Style:

  • Trigger words: "dwinking, can, bottle, glass, vial"

  • Use straightforward and descriptive phrases, such as:

    • "1 girl, blue shirt, red hair, looking at viewer"

• Quality Tags:

  • Avoid using quality indicators like:

    • "score_9"

    • "score_7"

• Important Note:

  • While the last two suggestions may still yield results, following these increases the likelihood of actually replicating the concept of drinking. It's also mostly due to how I auto-captioned these.

• I also uploaded the LoRA to CivitAI, so you can check out the gallery showcases there and see if any works well for you.