Totally incoherent!

#6
by A-Dragon-Speaking-Naturally - opened

Not one prompt put in generated anything close to useful.

Microsoft org

Hi,

The Text2Image foundation model of Visual ChatGPT is stable diffusion. It may occasionally generate some irrelevant images, but for most of the time it should be fine. Can you provide some examples to show where it fails?

-chenfei

I gave it an image of a beautiful woman in black armor with dark background and asked it to replace the background with a castle. It destroyed the woman's head in the process and when I asked it to return the head to its original form it generated a completely new photo of a rough drawn woman in armor telling me this was the original image. 75% of the time if you submit an input it just does nothing and requires you to repeat the input. At the very least stable diffusion does not just ignore your inputs.

Microsoft org

Could you open an issue in GitHub and paste a gif so we can help to locate the issue?

LanHarmony changed discussion status to closed

Sign up or log in to comment