hjunior29/video-text-remover 🖼️🔢❓ → 🖼️
About
Clean videos by automatically removing text overlays
Example Output
Output
Performance Metrics
137.95s
Prediction Time
155.94s
Total Time
All Input Parameters
{
"video": "https://replicate.delivery/pbxt/O8EwuGolBclczH4mMyfX18R6elbxm8KroY2y2cRdl3VvKboF/RONALDAO%20DO%20FUT.mp4",
"margin": 10,
"method": "hybrid",
"resolution": "720p",
"iou_threshold": 0.2,
"conf_threshold": 0.25,
"detection_interval": 0
}
Input Parameters
- video (required)
- Input video file with hardcoded text to remove. Supports MP4, AVI, MOV, and other common video formats.
- margin
- Extra pixels to expand around detected text regions (0-20). Higher values ensure complete removal but may remove more content. Recommended: 5
- method
- Video Text Remover removal algorithm. 'hybrid' (recommended): Best quality using context-aware inpainting. 'inpaint': Fast TELEA inpainting. 'inpaint_ns': Navier-Stokes inpainting. 'blur': Gaussian blur. 'black': Fill with black. 'background': Fill with surrounding color.
- resolution
- Processing resolution. Lower resolutions are significantly faster. Output will always be restored to original resolution.
- iou_threshold
- Intersection-over-Union threshold for removing duplicate detections (0.0-1.0). Higher values keep more overlapping boxes. Recommended: 0.45
- conf_threshold
- Detection confidence threshold (0.0-1.0). Lower values detect more text but may include false positives. Recommended: 0.25
- detection_interval
- Run text detection every N frames. Higher values speed up processing but may miss fast-appearing text. Recommended: 5
Output Schema
Output
Example Execution Logs
Processing video: /tmp/tmpy0e9sjooRONALDAO DO FUT.mp4 - Method: hybrid - Confidence: 0.25 - Margin: 10px - Processing resolution: 405x720 (720p) - Output will be restored to original resolution Video info: - Original Resolution: 2160x3840 - FPS: 30.00 - Total frames: 980 - Duration: 32.67s Processing frames with 128 workers (Batch size: 64)... Processing frames... Progress: 1/980 frames (0.1%) - Text detected in 64 frames Progress: 10/980 frames (1.0%) - Text detected in 64 frames Progress: 30/980 frames (3.1%) - Text detected in 64 frames Progress: 50/980 frames (5.1%) - Text detected in 64 frames Progress: 60/980 frames (6.1%) - Text detected in 64 frames Progress: 90/980 frames (9.2%) - Text detected in 128 frames Progress: 100/980 frames (10.2%) - Text detected in 128 frames Progress: 120/980 frames (12.2%) - Text detected in 128 frames Progress: 150/980 frames (15.3%) - Text detected in 192 frames Progress: 180/980 frames (18.4%) - Text detected in 192 frames Progress: 210/980 frames (21.4%) - Text detected in 256 frames Progress: 240/980 frames (24.5%) - Text detected in 256 frames Progress: 270/980 frames (27.6%) - Text detected in 320 frames Progress: 300/980 frames (30.6%) - Text detected in 320 frames Progress: 330/980 frames (33.7%) - Text detected in 384 frames Progress: 360/980 frames (36.7%) - Text detected in 384 frames Progress: 390/980 frames (39.8%) - Text detected in 445 frames Progress: 420/980 frames (42.9%) - Text detected in 445 frames Progress: 450/980 frames (45.9%) - Text detected in 509 frames Progress: 480/980 frames (49.0%) - Text detected in 509 frames Progress: 510/980 frames (52.0%) - Text detected in 509 frames Progress: 540/980 frames (55.1%) - Text detected in 573 frames Progress: 570/980 frames (58.2%) - Text detected in 573 frames Progress: 600/980 frames (61.2%) - Text detected in 637 frames Progress: 630/980 frames (64.3%) - Text detected in 637 frames Progress: 660/980 frames (67.3%) - Text detected in 701 frames Progress: 690/980 frames (70.4%) - Text detected in 701 frames Progress: 720/980 frames (73.5%) - Text detected in 765 frames Progress: 750/980 frames (76.5%) - Text detected in 765 frames Progress: 780/980 frames (79.6%) - Text detected in 823 frames Progress: 810/980 frames (82.7%) - Text detected in 823 frames Progress: 840/980 frames (85.7%) - Text detected in 882 frames Progress: 870/980 frames (88.8%) - Text detected in 882 frames Progress: 900/980 frames (91.8%) - Text detected in 944 frames Progress: 930/980 frames (94.9%) - Text detected in 944 frames Progress: 960/980 frames (98.0%) - Text detected in 944 frames Encoding video with FFmpeg... - Upscaling output back to 2160x3840... - Video encoded successfully with H.264 RESULTS: - Frames processed: 980 - Frames with text: 964 (98.4%) - Total text regions removed: 1916 - Average detections per frame: 1.96 Output: /tmp/tmpramlw3ck/output.mp4 Size: 14.00 MB
Version Details
- Version ID
247c8385f3c6c322110a6787bd2d257acc3a3d60b9ed7da1726a628f72a42c4d- Version Created
- November 28, 2025