The capacity of generative AI models such as ChatGPT and Gemini to produce images with remarkable quality continues to impress me, despite having witnessed numerous instances of how proficient these image generation models are. These models are capable of editing actual photographs by adhering to straightforward text prompts. Furthermore, these models can be exploited to create deceptive visuals that have the potential to become viral online.
The companies responsible for these innovations persist in launching enhanced text-to-image AI models, with Google being the most recent contributor. Google unveiled Imagen 4 at I/O 2025 last month, stating that the new model is considerably faster and superior to Imagen 3, which is already an advanced image generation service.
At that time, Google mentioned that Imagen 4 possesses “extraordinary clarity in intricate details such as complex fabrics, water droplets, and animal fur, and excels in both photorealistic and abstract representations.” Other enhancements include support for 2K resolution and refined spelling and typography.
Google also hinted that a rapid variant of Imagen 4 would be available shortly, offering a 10x speed boost compared to Imagen 3.
Currently, Google has declared that Imagen 4 and Imagen 4 Ultra are accessible to users, showcasing several astonishing examples of what the new models can accomplish.
Cost and availability
On Tuesday, Google stated in a blog post that Imagen 4 and Imagen 4 Ultra are now accessible to paid preview users in the Gemini API. For a limited duration, Google AI Studio will also facilitate Imagen 4 testing.
Imagen 4 will be priced at $0.04 per output image, one cent more than Imagen 3. Opt for the Ultra edition, and you’ll incur a fee of $0.06 per image. Additional tiers will be revealed in the future. All Imagen 4 images will include a non-visible SynthID watermark. However, a visible watermark will not be present.
