Google has announced the relaunch of its AI image generation capabilities, addressing previous issues and introducing new safeguards.
The company’s latest image model, Imagen 3, will be rolling out to users of its Gemini AI platform in the coming days.
Key Updates:
- Imagen 3 offers improved image quality and prompt understanding.
- New safeguards aim to prevent problematic outputs.
- Image generation of people to return, with limitations.
Addressing Past Controversies
In February, Google suspended its AI image generation feature after users reported historical inaccuracies in generated images.
Dave Citron, Senior Director of Product for Gemini, stated that Google has “worked to make technical improvements to the product, as well as improved evaluation sets, red-teaming exercises, and clear product principles.”
New Capabilities & Limitations
Google claims Imagen 3 can generate high-quality images across various styles, from photorealistic landscapes to stylized artwork. The model boasts improved text rendering and a better understanding of complex prompts.
However, Google has implemented the following restrictions:
- No generation of photorealistic, identifiable individuals
- No depictions of minors
- Restrictions on excessively violent, gory, or sexual content
Rollout Plans
The new image generation features will initially be available to Gemini Advanced, Business, and Enterprise users, starting with English language support. Google plans to gradually expand to more users and languages.
Other AI Developments
This relaunch comes alongside other Gemini updates, including:
- Custom “Gems” allowing users to create specialized AI assistants
- Expanded availability of Imagen 3 across Google’s AI products
Google’s careful approach to relaunching image generation highlights the challenges balancing innovation with responsible development.
However, the company admits it isn’t a solved problem, stating, “not every image Gemini creates will be perfect.”
