Try size: 768 x 768 or 512 x 768 (1024x 1024 or 768 x 1024 or above requires more sampling steps)

Weights: 1.0

Note: Please use natural language for complex elements, and it's okay to use short tags for simple elements.

Training data: If you wish to training other versions, you do not need my permission. As long as it is within reasonable and legal usage. Perhaps in the future, there will be base models better than Flux. And I might not necessarily still be on Civitai. The other reason for providing a Training Data is... if your descriptions or ChatGPT are not effective, these caption can serve as a reference. Or it can also be done through image analysis:

https://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha

Reference video: