Hello this is an experimental workflow that i made,i put images if you want to try yourself.Prompt : Change the car color to white,add custom car wrap black stripes on the car door,Make weather snowy,everything covered in snow,turn on the car lights,Change time to sunset,Here is the notes :The Primitive which is downsampling factor range from 0 to 5.The primitive value of 1 : Image becomes similiar to originalThe primitive value of 3 : Image follows prompt moreWeight value is also effective, weight 1 = similiar to original--------------Recommended and tested values--------------------------Primitive : 1.7 = Similiar to input image and prompt following but more similiar to input image.Primitive : 3.0 = More freedom, and more prompt following.Extra notes :I tried some loras with this kontext but it is not stable,couldn't find suitable value for lora strength,do not use any lora is recommended from my perspective.This minuszone node allows you to run this workflow on 6GB vram, it keeps loading the model partly, you can see at your vram usage.If you have 12GB + vram you can bypass.The link :<a target="_blank" rel="ugc" href="https://github.com/MinusZoneAI/ComfyUI-FluxExt-MZ">https://github.com/MinusZoneAI/ComfyUI-FluxExt-MZ</a>Thanks to t8star on youtube so that i could run this workflow withthis node on lowvram.credits : <a target="_blank" rel="ugc" href="https://www.youtube.com/@T8star-Aix">https://www.youtube.com/@T8star-Aix</a>I also started new youtube channel if you want to follow,i will maketutorial videos (i hope).my youtube : <a target="_blank" rel="ugc" href="https://www.youtube.com/@AITold/videos">https://www.youtube.com/@AITold/videos</a>This is experimental workflow,try it on your own values. The original black car image got from <a target="_blank" rel="ugc" href="http://pexels.com">pexels.com</a>

Files with SHA256: c5fefa5a262b0f3ade853fcee017226fcbe23620473676344617898a8186a20a