Combine with character lora?
Can we use this combined with a character lora to maintain character consistency between angle changes? Will a base qwen character be sufficient or do we need to train a 2509 character lora? What would we use for control image if training a character lora for 2509 if that is necessary?
Can we use this combined with a character lora to maintain character consistency between angle changes? Will a base qwen character be sufficient or do we need to train a 2509 character lora? What would we use for control image if training a character lora for 2509 if that is necessary?
I haven't tried it myself, but you can test it first. Use large images, medium images, and close-up images of the text to test whether the text is distorted or blurred. If distortion occurs in all of them, you might need to train a control LoRa. If only small images are distorted, then it's most likely due to the model's limitations.