Google recently announced the availability of its newest text-to-image models, "Imagen 4." Developers can access it as a paid preview via the Gemini API, with a limited free trial offered within Google AI Studio starting June 24, 2025.
The introduction of these models marks another addition to Google's suite of creative AI tools.
A Look at "Imagen 4" and "Imagen 4 Ultra": Capabilities and Pricing
"Imagen 4" can generate up to four 1024x1024 pixel images per request and accepts text prompts up to 480 tokens. Creating a single image costs approximately $0.04.
Alongside the base version, Google announced a more precise edition, "Imagen 4 Ultra." Priced at about $0.06 per image, it offers more accurate responses to complex text prompts.
According to Google, both models include invisible digital watermarks using "SynthID" technology to enhance transparency and provide near-instant generation speeds.
"Imagen 4 Ultra" boasts additional capabilities, including tighter alignment with user text prompts and the option to export images at 2K resolution via the Vertex AI platform.
How to Access and Usage Limits
For developers, access to these models will be through the same /generate endpoint used for Gemini models.
Google has indicated that higher usage limits can be requested during the preview period; the default limit is 20 requests per minute per project.
You can try the models for free via the following link:
https://aistudio.google.com/app/prompts/new_image
Initial Experiences: "Imagen 4" and "Ultra" Under Test
Early users have started sharing their experiences and comparisons on social media platforms. They have noted improvements in the clarity of text generated within images and reduced visual distortions compared to the previous "Imagen 3" version.
Practical experiments demonstrate the new models' capabilities. For instance, using a specific text prompt, the "Imagen 4" model successfully created a realistic depiction of a rainy night scene, showcasing a neon sign and its reflection.
Another experiment, targeting "Imagen 4 Ultra's" ability to generate more complex text within artistic contexts, used the following prompt:
"Imagen 4 Ultra" demonstrated its capacity to render this colorful and detailed scene, clearly highlighting the requested phrase amidst multiple graffiti designs.
These updates are part of Google's ongoing efforts to enhance its suite of AI-powered creative tools, which also includes the "Veo" models and the "Gemini" family of models.
By providing generous free usage tiers and a flexible usage-based pricing system via Google Cloud, the company aims to attract both developers and organizations.
Nevertheless, some observers suggest that the resulting images, despite their high quality and accuracy in following prompts, may sometimes lack a distinctive artistic touch or appear clearly machine-generated.
It remains to be seen how the creative and tech communities will receive these new tools, and whether they will indeed bring about a significant shift in AI image generation, especially as Google plans for a broader public release in the coming weeks.
