Imagen 4 Ultra T2I

Imagen 4 is a text-to-image model from Google that focuses on high visual fidelity, better prompt alignment, and improved rendering of text and fine details. It accepts longer prompts (up to 480 tokens) and can generate up to 4 images per request. The model supports multiple aspect ratios and resolutions including 1024×1024, 1792×2560, 2560×1792, 1536×2816, and 2816×1536. It incorporates watermarking via SynthID and includes safety controls to filter inappropriate content. The “Ultra” variant emphasizes stricter prompt adherence and precision in output, making it suitable for use cases requiring high visual accuracy. The model is available via Google’s Gemini API and limited preview in AI Studio.

June 24, 2025

Google

Text to Image (T2I)