According to reports, Adobe is developing an AI-powered text-to-video generation model and plans to enhance its capabilities by acquiring video content from photographers and artists. The company plans to use this content and its existing library of stock images and videos to effectively train its AI models. Adobe is offering an average fee of $2.62 (approximately 220 rupees) for each minute of video content submitted, according to sources cited by Bloomberg.
What is Adobe planning?
Adobe is reaching out to a network of photographers and artists and offering to pay them up to $120 (approximately Rs. 10,000) for video submissions. These videos have been requested to feature individuals engaged in everyday activities and expressing various emotions such as joy and anger. Adobe is apparently aiming to use this data to train its AI models to recognize and reproduce human facial expressions and natural movements.
In addition, Adobe requested specific types of video content, such as short clips depicting people expressing emotions and shots that focused on different aspects of the human anatomy, such as hands, feet, and eyes. Additionally, the company solicited videos of individuals interacting with objects such as smartphones or gym equipment.
The document, obtained by Bloomberg, emphasizes the importance of following guidelines regarding copyrighted material, nudity and offensive content when posting videos. The average compensation offered is $2.62 per minute, but some posters may receive up to $7.25 (approximately Rs. 600) for the content they post.
Rise of AI models
This development highlights a growing trend for companies to invest in sourcing data to train AI models as publicly available data sources become increasingly depleted. While some companies choose ethical data acquisition methods, others are facing criticism for allegedly using copyrighted material obtained from social media platforms. For example, a recent report claims that OpenAI utilized over 1 million hours of his transcription data from YouTube videos to train his GPT-4 model.
