- Adobe is leveraging its network of photographers and videographers to train its AI text-to-video generator.
- The company is acquiring diverse footage showcasing everyday activities and emotions.
- Compensation for video submissions raises questions about fairness, with rates averaging $3 per minute.
- Industry comparisons reveal varying pricing structures for AI training data.
- Adobe’s approach emphasizes exclusivity by utilizing proprietary in-house assets for AI model training.
- Competitors like OpenAI and Google resort to public data sources like YouTube transcripts.
- X’s updated privacy policy hints at utilizing user-generated content for AI training.
Main AI News:
Adobe, a titan in creative software, is venturing further into the realm of artificial intelligence with its latest project. In a bid to enhance its AI text-to-video generator, Adobe is tapping into its vast network of photographers and videographers. The company aims to curate a diverse array of footage capturing everyday human activities and emotions.
With competitors like Google’s Lumiere and OpenAI’s Sora setting high standards, Adobe is leaving no stone unturned. It’s acquiring footage portraying mundane tasks like walking, running, and using cell phones, alongside emotive scenes depicting sadness, excitement, or rage. Additionally, Adobe seeks simple yet crucial footage illustrating human anatomy, an area where AI often falters.
However, questions arise regarding Adobe’s compensation model. Bloomberg reports a payment of $120 for 40 to 45 minutes of content, averaging around $3 per minute. Despite the lack of creator recognition or royalties, is this a fair rate? This echoes Adobe’s previous call for still images, offering a meager 6 to 16 cents per image.
A glance at the industry landscape reveals varied pricing structures. Reuters cites rates ranging from 5 cents to $1 per photo and over $1 per video. Companies like Defined.ai broker AI training data, with prices ranging up to $2 per image and $2 to $4 per short video. Longer videos command even higher rates, from $100 to $300 per hour.
Since 2023, Adobe has been at the forefront of AI innovation, introducing products like Firefly, its text-to-image generator. Now, with AI-generated video on the horizon, Adobe is racing to stay ahead.
AI development hinges on vast datasets, often raising privacy concerns. Unlike competitors leveraging publicly available content, Adobe takes a distinct approach. It trains its AI models on proprietary in-house assets, crafted specifically for this purpose. This not only ensures exclusivity but also aligns with Adobe’s focus on marketing and creativity.
In contrast, reports suggest competitors like OpenAI and Google resort to YouTube transcripts for training data. Meanwhile, X’s updated privacy policy hints at leveraging user-generated content for AI training, highlighting the diverse approaches within the industry.
Conclusion:
Adobe’s strategic investment in AI video training signals a significant move in the market. By tapping into its vast network and prioritizing proprietary data sources, Adobe aims to set itself apart in the competitive landscape. However, the varying compensation rates and approaches within the industry underscore the complexities of AI development and data sourcing. As Adobe continues to innovate, its success will depend on navigating these challenges while staying ahead of the curve.
