When using the XLabs sampler in img2img the results retain a level of noise that looks similar to what you get out of the SamplerCustomAdvanced node with certain sampler/scheduler combos when comparing its output to denoised_output. The noise retention does not seem like an actual denoising issue during inference but rather something being suboptimal about the way it outputs the result.
This is an example of a 2nd Pass run with the XLabs Sampler and with SamplerCustomAdvanced
When using the XLabs sampler in img2img the results retain a level of noise that looks similar to what you get out of the
SamplerCustomAdvanced
node with certain sampler/scheduler combos when comparing itsoutput
todenoised_output
. The noise retention does not seem like an actual denoising issue during inference but rather something being suboptimal about the way it outputs the result.This is an example of a 2nd Pass run with the![Flux_sampler_compare](https://github.com/user-attachments/assets/4a18fe21-a1be-4a00-b6ce-bacdf4b6402c)
XLabs Sampler
and withSamplerCustomAdvanced