The proliferation of deepfakes has reached alarming levels, with a 4x increase in deepfakes worldwide from 2023 to 2024, according to ID verification platform Sumsub. In response, Meta has announced the release of Video Seal, an open-source tool designed to apply imperceptible watermarks to AI-generated video clips, helping to combat the growing menace of deepfakes.

Video Seal is the latest addition to Meta's arsenal of watermarking tools, which already includes Watermark Anything (re-released today under a permissive license) and Audio Seal. According to Pierre Fernandez, AI research scientist at Meta, the tool was developed to provide a more effective video watermarking solution, particularly for detecting AI-generated videos and protecting originality.

While Video Seal is not the first technology of its kind, Fernandez asserts that many existing approaches fall short. He cites limitations such as insufficient robustness to video compression, lack of efficiency, and non-reproducibility. In contrast, Video Seal is designed to be resilient against common edits like blurring and cropping, as well as popular compression algorithms. The tool can also add a hidden message to videos that can later be uncovered to determine their origins.

However, Fernandez acknowledges that Video Seal has certain limitations, mainly the trade-off between how perceptible the tool's watermarks are and their overall resilience to manipulation. Heavy compression and significant edits may alter the watermarks or render them unrecoverable. Despite these limitations, Meta is hopeful that Video Seal will contribute meaningfully to the fight against deepfakes.

To encourage widespread adoption of Video Seal, Meta is launching a public leaderboard, Meta Omni Seal Bench, dedicated to comparing the performance of various watermarking methods. The company is also organizing a workshop on watermarking this year at ICLR, a major AI conference. Fernandez expressed his desire to collaborate with the industry and the academic community to progress faster in the field, stating, "We hope that more and more AI researchers and developers will integrate some form of watermarking into their work."

The release of Video Seal comes at a critical time, as deepfakes continue to pose a significant threat to online authenticity. With the commoditization of generative AI, the production of fake content has become increasingly accessible, making it essential for technology companies and researchers to develop effective countermeasures. While Video Seal is not a silver bullet, it represents a significant step forward in the fight against deepfakes, and its open-source nature could facilitate further innovation and collaboration in the field.

In the broader context, the development of Video Seal highlights the need for a more concerted effort to address the challenges posed by deepfakes. As the technology continues to evolve, it is essential for stakeholders to work together to develop effective solutions that can keep pace with the rapidly changing landscape of online content. With Video Seal, Meta has taken a significant step in the right direction, and it remains to be seen how the industry and academic community will respond to this call to action.