RackFocus-Wan2.1-I2V-14B. A very interesting LoRA,it works well when the photos have out of focus blur. You can even fix the out of focus photos by generating the video with this LoRA. However, you may need to try some random seeds to generate the best result.You can use MediaInfo to check the generation info in the video.
UPDATED Wan2.2-High
The LoRA is for Wan2.2-High.Most of the image I test have a good result.
Strength:0.8-1.0, Prompt Suggestion:
rack focus
The Image below contains the Workflow.
Wan2.1
The workflow I used ComfyUI-WanVideoWrapper. I uploaded an article of my experience of training Wan2.1 LoRA (Both I2V & T2V): My Personally Training Experience of WAN:Starting with Data.
Description
FAQ
Comments (5)
I think this lora is ideal for moving focus like in movies and dramas. Is it possible to make a variation of this lora that will focus everything from the front to the back of the screen? For example, when I make a video with a close-up of face on wan, the background is always goes out of focus.
You mean just keep focus on the front and let the background blur instead of focus shift? Interesting, I will try.
_GhostInShell_
No, it's not. I'm not a native English speaker so I guess my expression was not accurate. Sorry. The variation I wrote is not a LORA with changing focus, but a Lora with everything in the screen from the foreground to the background in focus. In other words, a very deep depth of field LORA. A lora with blur removal. Here is a lora with a similar concept in FLUX. https://civitai.com/models/1737381?modelVersionId=1966280
fantaseed ok, I get it, let everything clear. I will try, I am not sure about the data
_GhostInShell_ It might be a data set you can generate with flux kontext? changing DoF/focus is something it can do