1. The workflow supports text-to-image generation. Disable the Depth module, and you can use reverse engineering of prompts or write them yourself.

2. Added the Depth Controlnet model based on FLUX text-to-image generation, which can reference the pose of the person in the image and use RH's official reverse engineering prompt nodes for images.

3. You can customize the F.1 Lora model, which allows generating portraits based on the pose of a reference image with your preferred Lora. Note that chaining multiple Loras may result in lower pixel quality.

4. Added a TTP image 2x upscaling feature, enhancing the details after upscaling.

5. Testing shows that image-to-image generation works quite well. Feel free to experiment more. Including the upscaling node, image-to-image generation usually takes about 4 minutes.