Google has launched ImageFX, an innovative AI-powered image creation tool that utilizes the Imagen 2 GenAI image model developed by Google’s DeepMind team. Similar to tools like OpenAI’s DALL-E 3 and Meta’s Imagine with Meta AI, ImageFX employs a prompt-based UI but introduces “expressive chips,” offering users keyword suggestions to explore different dimensions of their creations.
In a blog post, Google emphasizes ImageFX’s focus on experimentation and creativity, allowing users to easily modify images with text prompts and expressive chips. To address concerns about potential misuse, Google claims to have implemented technical safeguards, including filters for violent, offensive, and sexually explicit content. The tool also includes a prompt-level filter for “named people” and tags images with a SynthID digital watermark for added safety.
ImageFX is available in Google’s AI Test Kitchen, a web app for experimental AI projects. Additionally, Imagen 2 is being expanded across various Google products and services, such as the next-gen AI search experience and Vertex AI, Google’s managed AI services. Imagen 2 powers text-to-image capabilities in Google Ads and Duet AI in Workspace, and is now integrated into Google’s Search Generative Experience (SGE), offering users image generation tools based on prompts.
Google has also introduced Imagen 2 through an API in Vertex AI for Google Cloud customers, enabling the generation of high-quality images based on simple or complex prompts. The technology is also accessible through Bard, Google’s AI-driven chatbot.
While Google has not disclosed the training data for Imagen 2, the blog post mentions extensive measures taken to ensure safety, including adversarial testing and the use of SynthID watermarks for image identification in Google Search and Chrome. The post does not reveal the specifics of the data used to train Imagen 2, leaving questions about the use of publicly available or copyrighted data in the training process.
Read More On: Thestartupscoup.Com