Overview
- Gemini users can now modify both AI-created and uploaded images using natural language prompts for tasks like background changes, object replacements, and adding elements.
- The feature is powered by the Gemini 2.0 Flash model and initially piloted in Google AI Studio in March 2025 before its current global rollout.
- Edits include an invisible SynthID watermark to identify AI-modified images, with visible watermarking currently under testing to enhance safety measures.
- Google’s models are designed to refuse harmful or offensive editing requests, addressing concerns about deepfakes and misuse.
- The feature is accessible on both the Gemini mobile app and web platform, offering integrated text and image outputs for creative projects like illustrated stories.