Google's AI-powered Search now lets you create images from text
Google has announced that Search Generative Experience (SGE) users will now be able to create images directly from text prompts. Smart move considering Microsoft unveiled similar features for Bing AI in March by leveraging OpenAI’s DALL-E model.
The new feature is powered by Imagen AI models according to Google spokesperson Craig Ewer.
To use Google’s SGE, you need to be registered to the Google Search Lab program. After that all you need is to type your query into Google’s search bar and SGE will create images based on your prompt. Users can refine the text prompts to tweak each image.
The feature will also be available in Google Images - prompting users to generate images based on search queries.
In a blog post, Google's Hema Budaraju has said that they aim to roll out generative AI responsibly which means the new feature must adhere to Google’s prohibited use policy for generative AI. Additionally, all images will have metadata labelling them as AI-generated, with embedded watermarks.
Ewer also noted that at the moment, photorealistic faces and prompts that include notable individuals will not be allowed, and the tool is limited to those 18 and over.
In addition to images, Google is adding AI-generated text drafts from search queries. Users can choose long, short, or casual tones, and export them to Docs or Gmail if they choose.
When asked which models power the draft writing features, Ewer said that it was not one but several models.
“SGE is currently powered by a variety of LLMs. The models used in SGE are trained on both data from publicly available sources and on human supervision data,” he said.
Ever since its launch in May, Google has slowly been adding new features to SGE and the new image and draft creation features are definitely a step in the right direction.