JoyAI runs on Intel XPU with Modern_Enable_Intel_XPU, but LightX2V does not fully support this model (noise output)

[Tyr0727]

Problem Description
I integrated the JoyAI - Image -Edit model into the LightX2V framework and ran it on Intel XPU with the
Modern_Enable_Intel_XPU skill enabled.
The Intel XPU setup and the skill itself are working correctly — the pipeline runs end-to-end without any runtime errors.
However, the final generated images are pure noise, which indicates that LightX2V does not currently support JoyAI - Image -Edit’s model architecture / assumptions, rather than an issue with the XPU backend or the skill.

Confirmed Working Components

• Intel XPU backend works as expected
• Modern_Enable_Intel_XPU skill works correctly
• Model is loaded successfully
• Pipeline executes end-to-end without errors
• Output tensor has correct shape and dtype

Observed Issue

• Final generated images are pure noise
• No runtime error, NaNs, or Infs
• Behavior is consistent and reproducible

Since JoyAI - Image -Edit is currently not supported by LightX2V, we would like to request adding official support for the model in the LightX2V framework .