Google is Optimizing the Gemini Image Generation Feature Allowing Users to Select Areas for Fine-Tuning
Google has recently updated its Gemini application and upgraded its model to Gemini 1.5 Flash version, providing a longer context window and much faster response speed than the Gemini 1.0 Pro version, importantly, it is also free to use.
In addition to the model update, Google has also optimized the image generation feature of Gemini. Originally, generating images required continuously modifying the prompt to produce images that meet users' needs. To simplify this process, Google is experimenting with selection-based modifications.
After generating an image, if users are unsatisfied with certain areas, they can select the parts they wish to modify and describe what changes they would like to see. This way, Gemini can replace the content in the specified area without the need for complex prompts.
This feature is currently being tested in the Gemini for Android version 15.29.34.29. It was discovered during an APK Teardown, but users cannot directly use it yet, as Google might need more time for testing.
Switching to selection-based modifications is beneficial for both users and Google. For users, the biggest advantage is that generating and modifying images becomes more convenient. For Google, it can save server computational power because Gemini does not need to re-understand the entire image, only needing to modify the user-selected area.
The downside is that selection modifications can currently only be used on mobile devices, as the Gemini web version does not yet have this interactive method. The web version can only interact with Gemini through prompts, so if Google plans to add a similar feature to the Gemini web version, it may need to adjust the web version accordingly.