model_name stringclasses 6
values | media_type stringclasses 2
values | watermark_type stringclasses 6
values | default_position stringclasses 4
values | resolution_tested stringclasses 4
values | detection_method stringclasses 7
values | removal_compatibility stringclasses 7
values |
|---|---|---|---|---|---|---|
Dola AI | Video | Visible Text Overlay | Bottom-Right (87% Y / 78% X) | 1080p | Boundary Color Weighted Interpolation | Supported (Soft Fill / Blur / Crop) |
Dola AI | Image | Visible Text Overlay | Bottom-Right | 1024x1024 | Boundary Interpolation | Supported (Soft Fill) |
Google Veo 3 | Video | Alpha Blended Overlay | Bottom-Right | 1080p | Alpha Channel Matrix Inversion | Supported (De-blend) |
Google Gemini | Image | SynthID + Visible Badge | Bottom-Right | 1536x1536 | Synthetic Provenance Inversion | Supported (Visible Layer Only) |
Nano Banana | Image | Pattern Stamp | Corner Edge | 1024x1024 | Kernel Morphological Clean | Supported (Border Match) |
Runway Gen-3 | Video | Alpha Mark | Bottom-Right | 720p | Temporal Smoothing | Experimental |
Luma Dream Machine | Video | Visible Watermark | Bottom-Center | 1080p | Edge Crop / Interpolation | Supported |
AI-Generated Video & Image Watermark Spatial Dataset
A reference coordinate dataset mapping visible logo overlays, positional anchors, and aspect-ratio bounds across leading generative video and image models (Dola AI, Google Veo, Gemini, and Nano Banana).
Dataset Summary
Generative media platforms frequently place deterministic or semi-transparent visual badges on generated outputs. This dataset compiles bounding-box coordinates, average pixel offsets, and optimal client-side restoration algorithms (such as boundary interpolation, alpha inversion, and border cropping).
Models Analyzed & Benchmarks
| Platform / Model | Type | Typical Placement | Optimal Local Processing |
|---|---|---|---|
| Dola AI | Video / Image | Bottom-Right (87% Top, 78% Left) | Soft Boundary Fill & Bottom Crop |
| Google Veo | Video | Bottom-Right | Direct Inversion Matrix |
| Google Gemini | Image | Bottom-Right Corner | Pixel Mathematical Unblend |
| Nano Banana | Image | Corner Logo | Boundary Inpainting |
Live Browser Implementation
A pure client-side implementation of these algorithms (operating 100% locally in browser memory without server uploads) can be tested at:
- DigVibes Dola AI Watermark Remover - Complete browser video/image editor.
- DigVibes Creative Utilities Hub - Full suite of generative cleanup tools.
- Google Veo Video Remover - MP4 video de-blending engine.
Usage with Pandas
import pandas as pd
url = "[https://huggingface.co/datasets/chamee30/ai-generated-video-watermark-patterns/raw/main/data.csv](https://huggingface.co/datasets/chamee30/ai-generated-video-watermark-patterns/raw/main/data.csv)"
df = pd.read_csv(url)
print(df.head())
- Downloads last month
- 66