Google has presented Gemini 2.5 Flash Imageits new model of vanguard’s generation and editing, which includes capacities to combine several images in one maintaining the coherence of the characters that appear, as well as making specific transformations with natural language, all through the ‘app’ of Gemini Y Google AI Studio.
The technological giant launched its Gemini 2.5 Flash reasoning model in April of this year to VERTEX AI, designed as a hybrid option by allowing to activate or deactivate its reasoning capabilities to find the appropriate balance between the quality of its answers, cost and latency.
Now, in the face of offering a better image editing experience for both developers and users in the Gemini’s ‘app’, with higher quality results and “a more powerful creative control”, Google has presented its new Model of generation of GEMINI images 2.5 Flash Image.
Internally known as ‘Nanobanana’, this model allows you to combine several images in one, all maintaining the coherence of the characters or objects that are included. Also, it also allows specific transformations of images through natural language.
As the company has detailed in a statement in its blog for developers, these capacities have been implemented in the “mode of creation” of Google Ai Studio as a evidence. So that developers can remix or give life to their ideas with a single instruction in natural language and subsequently share it on Google Ai Studio or save the code in Github.
The native image edition in the application of Gemini for all users has also been updated, under the aim of improving editing capabilities and “maintaining the similarity between the images”, since “a similar but not the same representation is not right.”
/
It will be enough to share with Gemini a photo to edit concrete issues and “give it a unique touch.” For example, requesting that the person who appears in the photo appears in new scenarios, but maintaining their original appearance in each image. Gemini can also vary his outfits or professions, even reimagining the person at another time “without losing his identity.”
In the same way, users can also merge photos, sharing several images and unifying them in one to create a completely new scene. Thus, you can share the image of a dog and a person so that they appear in a cohesive way in the same image.
In addition to all this, concrete modifications of images can also be requested such as changing the color of the walls of a room or adding furniture, keeping the rest of the intact image.
Look: What is a “Roast Me” for the new trend in Chatgpt?
Google has also detailed other novelties offered by the model is the designs mixture. That is, apply the style of an image to an object of another. For example, use the color and texture of a flower petals for rain boots.
“This update does a much better job, allowing the editions to be more fluid, and the results of the model can be used for whatever you want,” said the product director in visual generation models in Google Deepmind, Nicole Brichtova, in statements to Techcrunch.
It also specified that all the images created or edited in the ‘app’ of Gemini include a visible water brand, as well as the invisible digital water mark Synthid.
However, Gemini 2.5 Flash Image is now available through the GEMINI and Google Ai Studio API for developers, as well as through VERTEX AI for companies. It must be taken into account that it has a cost of 30 dollars (25.86 euros per change) per million output tokens, and each image is equivalent to 1,290 output tokens. That is, the generation of an image has a cost of $ 0.039 (0.034 euros to change).
For its part, the update of the native image edition in the application of Gemini is now available for all users for free.