When we first started Vloggi five years ago, the dream was to use data to automatically compile videos from user-generated video sources based on rules-based workflows using the metadata contained within the mobile phone footage. Now that generative AI is a reality, our focus has shifted to providing the raw materials for the generative AI video world—footage.
In the age of generative AI video creation, companies suddenly need to acquire video assets at an unprecedented scale to create AI training datasets. For instance, every second of generative AI video requires over 1,000 videos to train the model. So, if a company has 1,000 product lines, it will require 30,000 video clips to produce one 30-second video or 300,000 clips to create a video for every product.
For every beautifully detailed and realistic video created by OpenAI’s Sora (above left) and its competitors, there are still hundreds of awful videos with extra limbs and illegible signs. Compare the two videos above, with exactly the same prompt, to give you some indication over the leap that we have to make before Gen AI is reliable enough for marketing videos.
In 2023, a video of actor Will Smith went viral, showing Gen AI’s limitations. The theory went that with sufficient video data sets with which to train a model, a famous actor would be easy for Runway to clone the gestures and facial expressions. The result was not quite that realistic. The problem was the pasta, not Will Smith.
So generative AI has a long way to go before it can be fully relied upon, but given the rapid advancements, this won’t take long. This is why companies are now in a race to stock their video libraries.
Why Video Asset Ownership is Crucial
Owning your video assets is crucial for several reasons:
- To Train Your Models: High-quality, proprietary video content is essential to train your machine learning models effectively. The richer and more diverse your dataset, the better your AI will perform. This diversity ensures the AI can handle various scenarios and nuances, leading to more accurate and reliable outcomes.
- To Prevent Competitors from Using Your Content: Controlling your video assets ensures that your competitors cannot use footage of your products to train their models, giving you a competitive edge. Ownership of unique content becomes a strategic asset, preventing others from leveraging your hard-earned media to enhance their offerings.
- To Combat Deepfake Fatigue: Using real people in your videos helps combat deepfake fatigue among consumers. Authentic, human-generated content resonates better and builds trust with your audience. As deepfake technology advances, the ability to distinguish genuine content from synthetic becomes crucial, and real human elements in videos can significantly enhance credibility.
This year, as a sign of the change, Vloggi has signed companies including Google and two US presidential candidates, as well as countless public advocacy bodies in the US, all eager to prove ownership of the material they are using. Our key selling point, one developed four years ago, is the digital content assignment form. That’s what we are now expanding on, to create a suite of UGC video content ownership tools.
Already, companies using Vloggi can download consent logs complete with IP addresses and a unique consent ID that is non-fungible. Additionally, companies can watermark all their videos and overlay contributor attribution to make it clear to would-be bad actors that the content is owned by the company and provided by the author.
Digital Consent Agreements
Digital consent agreements provide companies with legal certainty that they own the rights to the videos. This ensures that all content used is authorized and that there are no legal repercussions regarding usage rights, allowing businesses to operate with confidence and integrity.
Blockchain Video Traceability
By using blockchain technology, we ensure the traceability and transparency of video assets. Each video’s origin and modifications are recorded in an immutable ledger, providing an auditable trail that guarantees the authenticity and integrity of your content.
Object Analysis and AI Video Creation
Our advanced object analysis tools offer context around the video content for better understanding and usage. This technology can identify and tag objects within videos, providing metadata that enhances searchability and categorization, and allowing companies to leverage specific clips more effectively.
Digital Watermarking and Video Asset Ownership
Not only do we watermark the content, but we also watermark the metadata, ensuring secure and identifiable assets. This dual-layer watermarking helps in protecting the intellectual property and proving ownership, deterring unauthorized use and distribution.
Hidden Security Features and Deepfake Fatigue
We incorporate hidden security features that certify the authenticity of the source file. These features are embedded within the video file and can be used to verify the origin and integrity of the content, protecting against tampering and forgery. Using real people in your videos helps combat deepfake fatigue among consumers, enhancing credibility.
Meta and EXIF Data Analysis for AI Training Datasets
Our system provides a high degree of certainty that a video was created by a human on a mobile device, rather than by generative AI. By analyzing metadata and EXIF data, we can confirm the authenticity and context of the video creation process, adding another layer of trustworthiness.
Full Traceability for Contributors
We offer full traceability for contributors, allowing them to see how and where their content was used. This transparency fosters trust and encourages more contributions, as contributors feel valued and informed about the impact of their content.
We call this the human intelligence required for artificial intelligence. By owning and managing your video assets, you can ensure the quality, authenticity, and exclusivity of the content that powers your AI models. This not only enhances your technological capabilities but also strengthens your brand’s trust and credibility in an increasingly AI-driven world.
