Google, in response to Meta’s Make-A-Video video-generating AI, unveils Imagen Video, an AI system capable of generating video clips based on text prompts with a high degree of controllability and artistic styles, showcasing a significant leap in video synthesis capabilities compared to existing systems. The team behind Imagen Video explains how it creates a 128-frame video by generating 16 frames initially and predicting additional frames. Trained on vast video and image-text datasets, Imagen Video can produce a range of videos, including Van Gogh type paintings and watercolours, plus has an understanding of depth to create immersive drone flythroughs, demonstrating its ability to accommodate various aesthetics. This breakthrough pushes the boundaries of AI-generated video content, opening up exciting possibilities for creative applications.
Jasper, a startup focusing on its AI-content platform, secures $125m at a $1.5b valuation, while simultaneously acquiring AI startup Outwrite, known for its grammar and style checker used by over a million users worldwide. The fresh funds will fuel Jasper's expansion of core products, enhancing customer experience, and extending their technology to more applications. Additionally, the investment will facilitate the integration of Outwrite into Jasper's brand, streamlining their offerings. Leveraging AI, Jasper empowers users to create content for blog articles, social media posts, website copy, and beyond, allowing them to describe their requirements in natural language, whether it's for SEO-rich content or repurposing existing material with added context.
Shutterstock partnership continues with OpenAI to delve into AI-generated imagery, integrating DALL-E 2 image-generating AI system into its content for global Shutterstock users, with a launch planned in the near future. The collaboration also introduces a framework to ethically compensate artists contributing to AI model development, with potential royalties for their intellectual property usage. AI-generated imagery involves machine learning technology trained on visual data to respond to text-based prompts with corresponding pictures, and has showcased significant advancements recently. Shutterstock has opened a waitlist for users eager to access the upcoming integration of its content with OpenAI's DALL-E 2 image generator.
Earlier this year, Instagram initiated testing in the U.S. for an age verification program, aiming to authenticate users claiming to be 18 or older. Utilising video selfies analysed by an artificial intelligence system, the Meta-owned platform now intends to extend this program to India and Brazil, with further plans to roll it out in the U.K. and EU by the year-end. Through the program, users can upload a video of themselves, which Instagram's AI system examines to confirm their age. For this feature, Instagram partnered with Yoti, a U.K.-based identity startup. Once users complete the video selfie following on-screen instructions, Meta shares it with Yoti for verification, employing its specially trained AI technology.
Reddit has revealed its collaboration with Oterlu, a Swedish startup specialising in machine learning-powered content moderation tools. The Oterlu team will join Reddit's Safety team, focusing on developing native machine learning moderation models with the ability to swiftly and accurately detect harmful content in various languages. In addition to this, the team will work on crafting new safety tools for Reddit moderators. The partnership is a crucial step in Reddit's ongoing efforts to invest in and enhance its internal Safety team, responsible for overseeing content policy. Leveraging Oterlu's expertise in natural language processing, artificial intelligence models, and machine learning algorithms, Reddit aims to bolster its automated safety capabilities, ensuring a safer and more advanced platform for moderators and users alike.
Back to Blogs