Feature Summary
have the ability to make the low noise sampler or high noise sampler work using self refining video sampling
Detailed Description
here is the search : https://agwmon.github.io/self-refine-video/
this will be very useful if you want better results without additional training
well i use it in comfyui using SamplerSelfRefineVideo in comfyui-kjnodes
and because i only use around 4 steps i prefer to use it on the low noise model with the sigmas from high noise model
and keep the high noise with ksampler
this is a picture of how it works with me :
and here is the workflow
https://www.mediafire.com/file/c424jqitsb1vwdo/Wan2.2_i2v_00026_.mp4/file
.
note : i used :
DasiwaWAN22I2V14BLightspeed_boundbiteHighV10.safetensors
DasiwaWAN22I2V14BLightspeed_boundbiteLowV10.safetensors
Alternatives you considered
No response
Additional context
No response
Feature Summary
have the ability to make the low noise sampler or high noise sampler work using self refining video sampling
Detailed Description
here is the search : https://agwmon.github.io/self-refine-video/
this will be very useful if you want better results without additional training
well i use it in comfyui using SamplerSelfRefineVideo in comfyui-kjnodes
and because i only use around 4 steps i prefer to use it on the low noise model with the sigmas from high noise model
and keep the high noise with ksampler
this is a picture of how it works with me :
and here is the workflow
https://www.mediafire.com/file/c424jqitsb1vwdo/Wan2.2_i2v_00026_.mp4/file
.
note : i used :
DasiwaWAN22I2V14BLightspeed_boundbiteHighV10.safetensors
DasiwaWAN22I2V14BLightspeed_boundbiteLowV10.safetensors
Alternatives you considered
No response
Additional context
No response