search

Yuanshi9815 / OminiControl

A minimal and universal controller for FLUX.1.
699 stars 36 forks source link

How to keep the word on the subject correct in the generated image? #25

Open KeyaoZhao opened 4 days ago

KeyaoZhao commented 4 days ago

Great jod. And I test some images, find the word on the subject may wrong in the generated images. Do you have any improvement idea or suggestion? Thanks a lot.

Yuanshi9815 commented 4 days ago

Thank you for your question! Here are some key tips for getting better text accuracy in generated images:

  1. Image Preparation: Input images are automatically center-cropped and resized to 512x512 resolution.

  2. Prompt Writing: When writing prompts, it's best to refer to the subject using phrases like:

    • "this item", "the object", "it"
    • For example:
      • "A close up view of this item. It is placed on a wooden table."
      • "A young lady is wearing this shirt."

  3. Current Limitations: The model currently works best with objects rather than human subjects, as human data wasn't included in the training dataset.

Let us know if you have any other questions!