Remove Objects from Videos Easily
Remove Objects from Videos Easily
[FULL GUIDE]
᛫
Apr 27, 2025
᛫
by Mickmumpitz
Introduction
This workflow allows you to create clean plates by simply describing what should be removed using a prompt.
It then generates a video where the specified object or person has been removed, giving you a clean plate for further editing or VFX work.
Introduction
This workflow allows you to create clean plates by simply describing what should be removed using a prompt.
It then generates a video where the specified object or person has been removed, giving you a clean plate for further editing or VFX work.
🎨 Workflow Sections
🟨 Important Notes
⬜ Input / Output / Model Loaders
🟩 Prompt / Conditioning
🟪 ControlNets / Adapters
🟥 Latent / Sampling
Installation
Download the .json file and drag and drop it into your ComfyUI window.
Install the missing custom nodes via the manager and restart ComfyUI.
Download Models
VACE-Wan2.1-1.3B-Preview:
https://huggingface.co/ali-vilab/VACE-Wan2.1-1.3B-Preview/blob/main/diffusion_pytorch_model.safetensors
You can rename the model to "VACE-Wan2.1-1.3B-Preview" if you want.
📁 ComfyUI/models/diffusion_models
Wan2_1_VAE_bf16.safetensors:
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors
📁 ComfyUI/models/vae
umt5-xxl-enc-bf16.safetensors:
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/umt5-xxl-enc-bf16.safetensors
📁 ComfyUI/models/text_encoders
4x-ClearRealityV1.pth:
https://huggingface.co/skbhadra/ClearRealityV1/blob/main/4x-ClearRealityV1.pth
📁 ComfyUI/models/upscale_modelsFlorence-2-large:
This should download automatically.
🎨 Workflow Sections
🟨 Important Notes
⬜ Input / Output / Model Loaders
🟩 Prompt / Conditioning
🟪 ControlNets / Adapters
🟥 Latent / Sampling
Installation
Download the .json file and drag and drop it into your ComfyUI window.
Install the missing custom nodes via the manager and restart ComfyUI.
Download Models
VACE-Wan2.1-1.3B-Preview:
https://huggingface.co/ali-vilab/VACE-Wan2.1-1.3B-Preview/blob/main/diffusion_pytorch_model.safetensors
You can rename the model to "VACE-Wan2.1-1.3B-Preview" if you want.
📁 ComfyUI/models/diffusion_models
Wan2_1_VAE_bf16.safetensors:
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan2_1_VAE_bf16.safetensors
📁 ComfyUI/models/vae
umt5-xxl-enc-bf16.safetensors:
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/umt5-xxl-enc-bf16.safetensors
📁 ComfyUI/models/text_encoders
4x-ClearRealityV1.pth:
https://huggingface.co/skbhadra/ClearRealityV1/blob/main/4x-ClearRealityV1.pth
📁 ComfyUI/models/upscale_modelsFlorence-2-large:
This should download automatically.
You can find the WORKFLOWS & EXAMPLE FILES here:
Before You Begin: Thank you for considering supporting us! Since these workflows can be complex, we recommend testing the free versions first to ensure compatibility with your system. We cannot guarantee full compatibility with every system that's why we always provide the main functionalities for free! Please take a moment to read through the entire guide. If you encounter any issues:
|
Instructions
1. Prepare the Input
In the "Input" section, upload your video.
It’s recommended to load around 81 frames for optimal results.
2. Select Object for Removal
In this section, the person or object you want to remove is selected and tracked using a prompt. The workflow generates:
A mask that precisely covers the object.
An additional video where the area of the removed object appears as a white shape. This white area acts as an empty canvas for WanVideo, which will later fill it with the clean background.
How this section works:
Use "Florence2Run" to describe the person or object to be removed.
Adjust the area of the mask using "Grow Mask With Blur" to fully cover the target.
In the "Video Combine" node, you can check if the mask is large enough or still needs adjustment.
⚠️ Important: Do not use the "Blur Radius" setting here, as it can cause issues with Van Video and lead to unwanted results.
3. Create a Prompt
In the "Prompt" section, describe the scene as it should look without the object or person you want to remove.
Optionally, use the "Negative Prompt" to explicitly specify what should be removed.
4. WAN VIDEO
In the WanVideo Sampler node, you can try different seeds if you're not fully satisfied with the initial result.
Changing the seed creates slight variations, helping you find the best version.
5. Upscale + Stitch Clean Plate Area
In this section, the original video is used, but only the area where the object was removed is replaced with the newly generated clean plate.
You can fine-tune this area using the "Grow Mask with Blur" node:
Expand: Defines how much the mask should grow around the removed object.
Blur Radius: Controls how soft the edges of the mask are, ensuring a smooth and natural transition between the original footage and the cleaned area.
⚠️ Please note: The final image is reformatted to 720p because the ImageCompositeMasked node consumes a significant amount of RAM and can cause crashes. In our tests, we were able to stitch the clean plate area back into 4K footage on a machine with 128 GB of RAM, using around 88 GB of memory. Unsurprisingly, the same workflow caused crashes on a system with only 64 GB of RAM.
Done!
By following these steps, you can easily remove unwanted elements from your video and create a clean plate ready for compositing or further VFX work.
Instructions
1. Prepare the Input
In the "Input" section, upload your video.
It’s recommended to load around 81 frames for optimal results.
2. Select Object for Removal
In this section, the person or object you want to remove is selected and tracked using a prompt. The workflow generates:
A mask that precisely covers the object.
An additional video where the area of the removed object appears as a white shape. This white area acts as an empty canvas for WanVideo, which will later fill it with the clean background.
How this section works:
Use "Florence2Run" to describe the person or object to be removed.
Adjust the area of the mask using "Grow Mask With Blur" to fully cover the target.
In the "Video Combine" node, you can check if the mask is large enough or still needs adjustment.
⚠️ Important: Do not use the "Blur Radius" setting here, as it can cause issues with Van Video and lead to unwanted results.
3. Create a Prompt
In the "Prompt" section, describe the scene as it should look without the object or person you want to remove.
Optionally, use the "Negative Prompt" to explicitly specify what should be removed.
4. WAN VIDEO
In the WanVideo Sampler node, you can try different seeds if you're not fully satisfied with the initial result.
Changing the seed creates slight variations, helping you find the best version.
5. Upscale + Stitch Clean Plate Area
In this section, the original video is used, but only the area where the object was removed is replaced with the newly generated clean plate.
You can fine-tune this area using the "Grow Mask with Blur" node:
Expand: Defines how much the mask should grow around the removed object.
Blur Radius: Controls how soft the edges of the mask are, ensuring a smooth and natural transition between the original footage and the cleaned area.
⚠️ Please note: The final image is reformatted to 720p because the ImageCompositeMasked node consumes a significant amount of RAM and can cause crashes. In our tests, we were able to stitch the clean plate area back into 4K footage on a machine with 128 GB of RAM, using around 88 GB of memory. Unsurprisingly, the same workflow caused crashes on a system with only 64 GB of RAM.
Done!
By following these steps, you can easily remove unwanted elements from your video and create a clean plate ready for compositing or further VFX work.
© 2025 Mickmumpitz
© 2025 Mickmumpitz
© 2025 Mickmumpitz