Google upgrades the Gemini Chatbot with a new AI image model that gives users thinner photo editing, a step that is intended to cover Openai’s popular image tools and draw users from Chatgpt.
The update, called Gemini 2.5 Flash Image, began Tuesday to all users of the Gemini application, as well as developers through the Gemini API, Google Ai Studio and AI platforms.
Gemini’s new AI image model is designed to do more precise edits in the images – based on users’ natural language requests – while maintaining the consistency of faces, animals and other details, which is struggling with most competitive tools. For example, ask Chatgpt or Xai’s Grok to change someone’s shirt color in a photo and the result may include a deformed person or a spoiled background.
Google’s new tool has already drawn attention. In recent weeks, social media users muddy Above an impressive AI image processor on the Crowdsourced, LMarena rating platform. The model appeared to users anonymously under the nickname “Nano-Banana”.
Google says it’s behind the model (if it weren’t apparent already Of all banana -related tips), which is really the inherent image capacity within the flagship model Gemini 2.5 flash AI. Google says the image model is a state-of-the-art in Lmarena and elsewhere.
“We really push the visual quality forward, as well as the model’s ability to follow the instructions,” said Nicole Brichtova, a product that leads to visual production models on Google Deepmind, in an interview with TechCrunch.
“This update is doing a much better job by making processes more smoothly and model outings are used for whatever you want to use,” Brichtova said.
AI images models have become a critical battle soil for the Big Tech. When Openai started the GPT-4O image generator in March, led the use of chatgpt through the roof thanks to an AI frenzy created by the Ghibli Memes studio that, according to Openai Sam Altman CEO, left the company’s GPUs “melting. ”
To keep up with Openai and Google, Meta announced last week that it will grant AI Image Models license from Midjourney Startup. Meanwhile, the German unicorns -backed German laboratories continue to dominate the reference points with Flux AI icon models.
Perhaps the impressive AI image processor of Gemini can help Google close the user’s gap with Openai. Chatgpt now records more than 700 million weekly users. About the call of Google profits in July, Tech CEO Giant Sundar Pichai revealed that Gemini had 450 million monthly Users – implying that weekly users are even lower.
Brichtova says Google has specifically designed the image model with cases of consumer use in mind, such as helping users to visualize their home and garden projects. The model also has a better “global knowledge” and can combine multiple references to a single exhortation. For example, merging a picture of a sofa, a living room photo and a color palette in a cohesive performance.
While Gemini’s new AI image generator makes it easier for users to make and process realistic images, the company has securities that limit what users can create. Google has struggled with the AI image generator safeguards in the past. At one point, the company apologized for the Gemini who create historically inaccurate images of people and completely turned the AI image generator completely.
Now, Google thinks it has hit a better balance.
“We want to give users a creative control so that they can get what they want from the models,” Brichtova said. “But it’s like nothing.”
The AI genetic section of Google’s Terms of Services forbid users from creating “non -conspiracy familiar images”. These same types of safeguards do not seem to exist for Grok, which allowed users to create AI clear pictures They look like celebrities, such as Taylor Swift.
To tackle the rise of Deepfake images, which may find it difficult for users to discern what is real online, Brichtova says that Google applies visual watercolors to AI -produced images, as well as identifiers in its metadata. However, a rolling beyond a picture of the social media may not seek such identifiers.
