Author: ShermanVV
Workflow Source: https://openart.ai/workflows/pika_creamy_57/flux pulid in context lora 2k out per image/JxsK7Y80WlY0sj0Ylo2i
Features
1. Use Flux Pulid In Context LoRA to reference a facial photo and achieve consistent output of the specified character.
Process
1. Load the model flux controlnet HD and enhance details with upscaling.
2. Crop into 4 images, and enhance details with upscaling for each image.
3. During the detail enhancement process, since the pulid model and reference images are involved, the final output images can maintain stable similarity.
Parameter Adjustment Instructions
1. In the separate input area, you can customize configurations such as:
Positive prompts,
Negative prompts,
Similarity,
HD strength,
Upscale redraw strength,
Upscale multiplier,
and other parameters, which can generally be fine-tuned.
2. The template for prompts includes descriptions of the entire image as well as detailed descriptions for the individual 4 images. Use tags [] to specify and include the following tags:
[FOUR PANEL]
[TOP LEFT]
[TOP RIGHT]
[BOTTOM LEFT]
[BOTTOM RIGHT]
Under each tag, describe the prompts for each image respectively.

3. Another point to note is that during the process of enhancing details after splitting the images, the prompts for each image also need to be configured correspondingly.

4. This workflow was run in an environment with a 4090 GPU and 24G VRAM; other environments have not been tested yet.