Using Regional Prompter Extension in Automatic1111

In this article, I aim to document my experiences using the Regional Prompter extension in automatic1111, a tool that enhances image generation by applying prompts to specific regions of the desired image. This helps greatly with composition. It is not 100% accurate but comes very close most of the time.

This isn't a tutorial on how to use Regional Prompter; rather, it's a compilation of useful tips and tricks that I've gathered.

For a comprehensive guide on Regional Prompter, visit this excellent resource: Stable Diffusion Art - Regional Prompter.

Comparison with Latent Couple Extension

The Regional Prompter seems to share functionality with another extension known as Latent Couple. I haven’t personally used Latent Couple, so your experience may vary.

Understanding Common Prompts Versus Base Prompts

Both common and base prompts are integral to Regional Prompter. Both prompts are applied to the overall image, but base prompts allow for more flexibility in how the prompt is applied to the regions created with Region Prompter.

• Common Prompts: These apply a uniform prompt across all image regions, streamlining the process and minimizing errors or duplication.

  • For instance, instead of repeatedly specifying details for each character in an image, you can simplify the prompts as follows:

    • Instead of: a man and woman, a man with black hair BREAK a man and woman, a woman with red hair

    • Use: a man and woman BREAK a man with red hair BREAK a woman with blue hair

• Base Prompts: Similar to common prompts, base prompts let you adjust their influence on the overall image using a base ratio. This feature provides control over how uniformly the prompt applies across different regions and allows for varying weights in targeted areas.

Using Multiple Character LoRAs with Regional Prompter

Characters often blend when using multiple LoRAs representing different characters in the same image. Regional Prompter allows for distinct character representation in the same image.

However, there are important considerations:

• Use Latent mode, not Attention mode

• If your generated images are degraded it may be because your LoRAs are corrupting the resulting image. There are some suggested fixes in the documentation. I've found lowering the weights of the LoRA and using ADetailer at a higher weight to fix the faces works well.

Integrating ADetailer with Regional Prompter

Regional Prompter can be effectively combined with ADetailer, especially when dealing with multiple faces in an image.

Use the [SEP] keyword to separate faces and the [SKIP] keyword to omit a face. The processing order of faces can be adjusted in the settings. Read this Reddit post for more information.

doctor strange [SEP] black widow