ayushunleashed/minimax-remover 🔢🖼️ → 🖼️

▶️ 165 runs 📅 Jun 2025 ⚙️ Cog 0.14.7 🔗 GitHub
video-editing video-inpainting video-object-removal

About

Remove any object from video - fast.

Example Output

Output

Performance Metrics

4.92s Prediction Time
124.17s Total Time
All Input Parameters 
{
  "fps": -1,
  "mask": "https://replicate.delivery/pbxt/JgGNe7IjsavH2is00AVgbEwWroEtseid6f3HZIHI7MJH9Num/bmx-trees-mask4.mp4",
  "video": "https://replicate.delivery/pbxt/JgGNdl0r9V14YxgJ4JlsCz71rSU9y8NJD9RDDdJo2Ua3r8C7/bmx-trees2.mp4",
  "width": -1,
  "height": -1,
  "num_frames": -1,
  "num_inference_steps": 6,
  "mask_dilation_iterations": 8
}
Input Parameters
fps Type: integerDefault: -1Range: -1 - ∞
Output video FPS (-1 = same as original video)
mask (required) Type: string
Mask video file where white areas indicate objects to remove. See examples: https://replicate.com/ayushunleashed/minimax-remover/readme
seed Type: integer
Random seed for reproducible results (leave blank for random)
video (required) Type: string
Input video file with objects to be removed
width Type: integerDefault: -1Range: -1 - ∞
Output video width (-1 = same as original video, auto-scaled to max 1920px if needed)
height Type: integerDefault: -1Range: -1 - ∞
Output video height (-1 = same as original video, auto-scaled to max 1920px if needed)
num_frames Type: integerDefault: -1Range: -1 - ∞
Number of frames to process (-1 = same as original video)
num_inference_steps Type: integerDefault: 6Range: 1 - 50
Number of denoising steps (higher = better quality, slower. 6=fast, 8=balanced, 12=high quality)
mask_dilation_iterations Type: integerDefault: 8Range: 1 - 20
Mask expansion iterations for robust removal (higher = more thorough removal)
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Using seed: 6788
Validating input videos...
📹 Processing 80 frames at 432x240
🎬 Output FPS: 24 (original: 24.0)
⚙️ Quality: 6 inference steps
🎯 Mask dilation: 8 iterations
Loading original video and mask...
Loaded 80 frames from 80 total frames
Loaded 80 mask frames
Video shape: torch.Size([80, 240, 432, 3])
Mask shape: torch.Size([80, 240, 432, 1])
Running MiniMax-Remover inference...
  0%|          | 0/6 [00:00<?, ?it/s]
 17%|█▋        | 1/6 [00:00<00:02,  2.30it/s]
 33%|███▎      | 2/6 [00:00<00:01,  2.94it/s]
 50%|█████     | 3/6 [00:00<00:00,  3.15it/s]
 67%|██████▋   | 4/6 [00:01<00:00,  3.34it/s]
 83%|████████▎ | 5/6 [00:01<00:00,  3.45it/s]
100%|██████████| 6/6 [00:01<00:00,  3.52it/s]
100%|██████████| 6/6 [00:01<00:00,  3.30it/s]
Inference completed successfully!
It is recommended to use `export_to_video` with `imageio` and `imageio-ffmpeg` as a backend.
These libraries are not present in your environment. Attempting to use legacy OpenCV backend to export video.
Support for the OpenCV backend will be deprecated in a future Diffusers version
✅ Video saved to: /tmp/tmpril9jgvo/output.mp4
📊 Output: 77 frames at 24 FPS
Version Details
Version ID
f486e4c1469ce3d5042ad74ab06b16138ae228ddb49b54ce050a14cb2f1beb58
Version Created
June 15, 2025
Run on Replicate →