The generated image is a bit blurry
As the first ipadapter model of Flux, it is very powerful. I tested it yesterday and the generated image was a bit blurry, like the feeling of not being fully generated. Can this issue be fixed?
As the first ipadapter model of Flux, it is very powerful. I tested it yesterday and the generated image was a bit blurry, like the feeling of not being fully generated. Can this issue be fixed?
So one thing that can definitely help is passing a high quality image as input, we trained the model on high quality images and when it is fed low resolution, low quality or blurry images it can really mess up the output quality, we are working on it now but for now we can recommend providing high quality images as input.
By the way, thanks for the feedback.
Thank you very much for your reply. My usage steps are as follows:
- A high-definition woman was used as a reference image
- Tested 20 steps first, there were some ambiguities
- Revised to 30 steps, with some improvement.
- Modify to 50 steps, with slight changes, but still some blurriness, as if there is a thin veil covering it.
original image
20step
30step
50step