Imagen 3 v2
Imagen 3 v2 is a generally available release of Google's Imagen 3 text-to-image model, producing photorealistic images with reliable text rendering and few artifacts. It adds a prompt-rewriting enhancement over the initial Imagen 3 release while keeping the same image quality, watermarking, and aspect-ratio support. The model generates images at resolutions up to about 1K, in aspect ratios including 1:1, 3:4, 4:3, 9:16, and 16:9, and embeds SynthID, Google DeepMind's imperceptible watermark, in every output for provenance tracking.
Key info
Input
Output
Features
Context window
—
Max output
—
Available routes
No routes currently available — Imagen 3 v2 isn't routed through the Opper gateway right now. It may return.
Contact us about this model →