OpenAI recently introduced its video generation AI tool called "Sora" for ChatGPT Plus and Pro users. But isn't the first-of-its-kind tool, as several AI tools can do the same trickery using AI. As the market for AI video generation is growing every day, it may soon become challenging to detect fake or AI-generated videos online. To combat this, Meta has introduced its watermarking tool called "Meta Video Seal." This tool applies an invisible watermark to AI-generated videos.
According to Meta, the new neural video watermarking tool "Video Seal," adds an invisible watermark to an AI-generated video that can be embedded into a video without affecting how they look. Video Seal can also add a hidden message to videos, which can later be detected to trace the source of the video. What makes "Meta Video Seal" special is its ability to withstand common video editing efforts including blurring, cropping, or compression algorithms for easier sharing online. This means that if someone edits the video containing the watermark by Meta Video Seal, the watermark will stay intact, helping verify its authenticity and source.
Meta is sharing this tool under a permissive license, alongside a research paper, training, and inference code. There is also an interactive demo available for anyone to try out the model. Meta says that "while AI tools can help bring the world closer together, it’s important that we implement safeguards to mitigate the risks of imitation, manipulation, and other forms of misuse that can undermine their benefits. Post-hoc watermarking is a crucial step towards better traceability for content and AI models."
With Video Seal, Meta aims to make it difficult for bad actors to misuse AI to speak fake content and information. While AI can do amazing things and make things easier for us, it is also important to have certain guardrails to ensure correct information is available to everyone.
Source: Meta
1 Comment - Add comment