Gemini 2.5 Flash Image
Generate, transform and edit images with simple text prompts, or combine multiple images to create something new. All in Gemini.
Keep characters consistent
Reuse the same characters while changing their outfits, poses, the lighting, or the scene. Or reimagine yourself – across decades, in different places, or in your childhood dream job.
Prompt, combine, create
Merge up to three images to create something new. Generate surrealist art, combine disparate photo elements, or seamlessly blend objects, colors, and textures.
Control the details
Create and edit images with powerful control. Replace the background, restore faded images, and change characters’ outfits. Keep tweaking until you’re happy, all with natural language.
Push design boundaries
Experiment with creative directions, or bring them into different contexts. Apply specific patterns to visible surfaces, or test out colors for fashion, design, and interior decoration.
One prompt, many possibilities
Generate multiple images using just one prompt to explore different creative avenues. Or create several images that work together to tell a complete story.
Key features
-
Multimodal understanding
Upload images and share text instructions with Gemini to create complex and detailed images.
-
Conversational inputs
Use everyday language while creating images, and keep the conversation going to refine what the model generates.
-
Real-world knowledge
Generate images that follow real-world logic, thanks to Gemini’s advanced reasoning capabilities.
Benchmarks
Gemini 2.5 Flash Image is a state-of-the-art image generation and editing model, with lower latency compared to other leading models.
Gemini 2.5 Flash Image was tested on LMArena as nano-banana.
Limitations
While Gemini can now create a wide range of images, we’re still working on improving key capabilities.
-
Factual representation
Not every image Gemini generates will be perfect – it can still struggle with small faces, accurate spelling, and fine details in images.
-
Character features
The model excels at character consistency, but it may not always get it right. We're working to make this consistency even more reliable.
Safety & Responsibility
We use extensive filtering and data labeling to minimize harmful content in datasets and reduce the likelihood of harmful outputs. We also conduct red teaming and evaluations on content safety, including child safety, and representation.
Image generation in Gemini has all our latest privacy and safety features. This includes SynthID, our tool that embeds an invisible digital watermark directly into an image, allowing it to be identified as AI generated.