Unique Images

IP adapter in comfui. I'll share a link to this workflow in the comments. I've loaded these two copyright free images. You can find these kind of images from free stock image websites com. Here I'm basically looking for images that don't have hands in the frame, as AI is not yet perfect with generated hands. So here I've downloaded these three free images. So here I've loaded these three images for the reference and hopefully this should give us a unique picture and with a unique face. I've used the apply IP adapter from encoded node, which takes reference images from this node encode IP adapter. This way we can adjust the weights to these images according to our preference to have a better control over the generation. For the checkpoint I'm using this realistic stock photo, which is an SDXL model. Make sure to use SDXL model here because we are using SDXL checkpoint. I've loaded the LCM Lora for faster generation. Let'set the resolution of the empty latent to 1024 for SDXL. We'll use batch size to 2, which will generate two images to see which one I Now since we are using an LCM Lora node, we'll set the sampling steps to only 8, CFG 1. 5 and set the sampler to LCM. Let's write the text prompt, Indian Bride, pose, and a negative prompt. Okay, let's run this. Oh, let's connect the clip to the model. So here's our images. It has taken most details from first and third image. That's because I've set a lower weight to the send image. The face is unique. We try again. Changing the weights, I'll use the face restore node to make a detailed and unique face. Let's run it again. We'll use only two images now. Let's run again. Okay, so this time it took more details from the send image according to the weights. So overall the image looks good and unique. The face is unique. Some issues with the hand here which can be fixed, but you can always chose the best one. So that's it for this video. Please support by subscribing. It will help to make more videos