This v2v LoRA removes anything obscuring the scene. It was made to "reconstruct" ground truth video from sparse data (when details of the scene can only partially be seen or the scene has semi-occlusion like smoke or haze.) If the data is completely invisible from any of the frames then it will have to be hallucinated, but if it is present anywhere, even semi-transparently, then it will be retained. Adjust the strength of the LoRA to remove more layers or go "deeper" into the scene. Start at a strength of about 1.25 and adjust to your liking. More complex removal requires higher strength.
Example prompt: "Remove the green plant from the foreground." is all that is needed to perform the task, but if details of the scene need to be hallucinated, it will be best to describe the scene in the background.
LTX still struggles with fast movement, but if you interpolate the input to 50fps and run inference at 50fps then it should help with fast moving scenes.
See examples for what this LoRA can do. >>Example Workflow Here<<
The split LTX models in the workflow can be downloaded from Kijai's page: https://huggingface.co/Kijai/LTX2.3_comfy/tree/main
Description
FAQ
Comments (23)
damn that's cool
Seriously, the knife removal got me.
another amazing lora concept. this plus EditAnything...I hope we get more people doing this type of stuff to be able to remove stuff easily.
EditAnything struggles sometimes to remove objects.
What a great concept! I’m excited to put this into practice. Thank you for all your hard work.
wow, im just stunned. how did u train this? did u have before & afters of difficult clean up shots?
Workflow returns error 'Tensors must have same number of dimensions: got 5 and 3'
Explain cpecific requirements. Is it require specific ltx model to function?
Strangely enough! In my tests, I can't remove anything from the video. I wonder if it's because this LoRA doesn't support GGUF distilled models?
have you tried increasing the strength of the lora to something like 1.6?
@WepeNerd No! I thought the LoRA was working fine at 1.25, so I didn't increase the LoRA strength. Are you saying it's not working because the LoRA strength isn't high enough?
@k3soft it depends on what is being removed, but sometimes it requires the strength to be pushed a lot, depending on the scene
@WepeNerd I initially tried to remove the guardrails from the scene, but didn’t get the expected results. Then I tried removing the watermark and got the same outcome. Finally, I tried removing the person from the scene, but the result was still the same
Please guys, share some of your results and tell us if this lora works with gguf workflows...this lora seems to be one of the smartest ever made for video editing.
This is amazing! It’s working perfectly for my V2V workflow and does a great job of processing the sections I need to delete.
good to know, in my case I struggle to create the results I want. For example I want to remove my hand from holding a pencil that writes, and it removes the pencil as well
Sorcery
This lora has huge potential 🤩 though I struggle with certain results. I guess we need a V2 that understands words like 'inside' ' above' 'between' 'behind' etc. and remove the 'forward' limitation. Imagine a figure dancing and you want to remove the person to leave the clothes only. If I'm missing something, please advice ✨
this can't be real
i hope it works on wgp
it does, just need to use 1 phase and set strength to 1.25. 1.5 is way too high and colours bleed, 1.0 is too low and no effect.
I am wildly impressed by this. I have to look up V2V training to see how to do it. This seems so powerful.
Dont work here just ouput the same video
Did you try strength 1.25? For me that was the only one that worked. Any lower and it wont work.
does not work
Amazing lora, at least it looks to be that way, I wasn't able to remove anything yet :) Still trying.