Google Imagen 4 / Ultra Launches on AI Studio!
Google launches Imagen 4 & Ultra on AI Studio with MCP/Jules integration. Test results, pricing, and creative experiments included.
"AI Disruption" Publication 6900 Subscriptions 20% Discount Offer Link.
Google has released its latest text-to-image model, Imagen 4, along with its Pro Max version Imagen 4 Ultra, to AI Studio and API.
You can try it for free with limited usage in Google AI Studio and pay for usage beyond the quota. Actual testing shows the free quota is quite generous - I ran over 40 images and it's still working.
Imagen 4 can generate up to four 1024 × 1024 images per call, with prompts supporting up to 480 tokens.
Price-wise, regular Imagen 4 costs about 4 cents per image, while the high-fidelity Imagen 4 Ultra costs about 6 cents per image.
Both versions include SynthID watermarks and generate at near-real-time speed. The Ultra version reportedly has better prompt understanding and can export 2K resolution images through Vertex AI.
Developers use the same /generate interface as Gemini models, with the ability to apply for higher rate limits during preview (default is 20 requests per minute per project).
Imagen 4: Regular vs Ultra Version
Imagen 4 (Regular): Daily workhorse. This is their flagship text-to-image model that handles most image generation tasks. Compared to Imagen 3, quality improvement is significant, especially in text generation capabilities.
Imagen 4 Ultra: Precision-focused. If you need images to strictly follow your instructions, you'll need the Ultra version. It generates images with higher prompt adherence and performs very competitively against other mainstream text-to-image models.