Vertex AI's Imagen 4 Models | Generated by AI
Based on the code snippet you provided, you are using the genai.Client to interact with a model for image generation on Vertex AI. The model you’ve specified is "imagen-4.0-generate-preview-06-06".
Here’s a breakdown of what’s new and what this model represents:
The Latest Image Generation Models on Vertex AI
The model you are using, imagen-4.0-generate-preview-06-06, is part of the latest Imagen 4 model family on Vertex AI. This model family also includes:
imagen-4.0-generate-preview-06-06: This is the standard Imagen 4 model, offering a balance of speed and quality.imagen-4.0-ultra-generate-preview-06-06: This “Ultra” version is designed for higher quality but is slower and more expensive.imagen-4.0-fast-generate-preview-06-06: This “Fast” variant prioritizes speed and lower latency over the highest possible quality.
These Imagen 4 models are a significant update, building on previous versions like Imagen 3 and Imagen 2. The key features and updates with Imagen 4 include:
- Improved Image Quality: The new models are designed to generate images with higher quality compared to their predecessors.
- Wider Aspect Ratio Support: Imagen 4 supports several aspect ratios, including
1:1,3:4,4:3,9:16, and16:9, giving you more control over the output. - Multilingual Prompting: While English is the primary language, these models also support text prompts in several other languages, including Chinese (simplified and traditional), Hindi, Japanese, Korean, Portuguese, and Spanish (in preview).
- Safety Features: The models include user-configurable safety settings and incorporate digital watermarking (SynthID) to identify AI-generated content.
The person_generation and safety_filter_level parameters in your code snippet are still relevant and allow you to configure the model’s behavior for responsible AI, as you have done with person_generation="ALLOW_ADULT" and safety_filter_level="BLOCK_LOW_AND_ABOVE".
Beyond Image Generation
In addition to the updated Imagen models, Google Cloud has also recently introduced other new generative AI models on Vertex AI, including:
- Veo 3 and Veo 3 Fast: These are the latest video generation models, allowing you to create high-quality videos from text or image prompts. Veo 3 is for high-quality, professional videos with features like native audio and lip-syncing, while Veo 3 Fast is optimized for speed.
- Gemini 2.5 models: These are the latest versions of the Gemini family of models, including Gemini 2.5 Pro, Gemini 2.5 Flash, and Gemini 2.5 Flash-Lite, offering advanced reasoning, multi-modality, and improved performance for various tasks.