Meta launches AI image generator ‘Imagine with Meta’

Meta has introduced a new standalone generative AI tool called "Imagine with Meta," enabling users to create images by describing them in natural language. Similar to OpenAI's DALL-E, Meta's tool uses the existing Emu image generation model to produce high-resolution images from text prompts. Currently free for users in the US, it generates four images per prompt and is accessible both within chats and on the web.

With ‘Imagine with Meta’ company aims to add more racial safeguards which had been a grave concern with Meta’s previous AI sticker generator. Meta is also planning to introduce invisible watermarks to the generated content in the coming weeks to enhance transparency and traceability. These watermarks, generated by an AI model, will be detectable with a corresponding model and are designed to withstand common image manipulations like cropping, resizing, and colour changes.

In the official announcement, Meta acknowledges the importance of making AI-generated content clearer, especially in light of concerns surrounding Deepfakes and AI-generated content bypassing filters. 

The central government has also expressed its concerns regarding the misuse of deep fake technology, directing the social media platforms to build mechanism to deal with such content. 

China's Cyberspace Administration has already implemented regulations requiring generative AI vendors to mark their content without affecting user usage.

In the US, Senator Kyrsten Sinema also underlined need for transparency in generative AI during recent Senate committee hearings, advocating for the use of watermarks. Meta's move to include watermarks in its AI-generated content aligns with these emerging standards and reflects a broader industry trend toward increased accountability in the use of generative AI.