Prepare and test your skills
Prepare and test your skills
Imagen (opens in a new tab) is Google's foundation model for high-quality image generation and editing. Imagen creates new high-fidelity images from text prompts and transforms existing images with guided editing modes. The core workflow moves from a text description to a new image, and from a source image plus a text description to an edited image.
Text-to-image generation is Imagen's primary function for creating new, high-fidelity images from a written prompt. A user provides a text description of the desired scene, object, or style, and Imagen generates original image content that matches that description. This approach supports creative production, marketing visuals, and concept exploration without starting from an existing photo or illustration. The quality of the result depends on the clarity and detail of the prompt.
Inpainting is Imagen's editing capability that changes content inside the boundaries of an existing image. The source image stays as the base, and Imagen regenerates only the targeted area according to the prompt. This flow moves from the original image plus a text instruction to an updated image where only the selected region is changed, while the rest of the picture is preserved.
Primary inpainting functions include:
Outpainting is Imagen's editing capability that expands image content beyond the original edges. The original image provides the visual context, and Imagen generates a natural continuation that extends the scene outward. Image-to-image transformation is the broader pattern behind all Imagen edits, where an input image plus a text prompt produces a new output image. This pattern connects inpainting, background replacement, and outpainting, because each starts with source pixels and uses language to direct the change.
Exam tip: Inpainting edits content inside the image, such as inserting or removing an object, while outpainting expands content outside the original image boundaries.
Strategic Business Use Cases and Integration for Imagen means using a foundation model for high-quality image creation to solve business content needs, and running it on managed enterprise infrastructure. Imagen is Google Cloud's foundation model (FM) for generating images, offered through Vertex AI (opens in a new tab) with generative AI tools and APIs. Business teams send text or image inputs to Imagen on Vertex AI, and Vertex AI returns new or edited images that can flow directly into marketing, commerce, and content workflows.
Core Imagen capabilities for business use are the Generally Available (GA) image functions that map to different content tasks. Text-to-image generation creates a new image from a text prompt, while image editing modifies an existing image with a prompt. Image visual captioning generates a text description of an image, and Visual Question Answering (VQA) answers questions about image content.
The GA capabilities include:
Human face generation is supported for image generation and image editing and is enabled by default, except for images with children and celebrities. Together, generation and editing produce visual content, while tuning, captioning, and VQA adapt and describe that content for consistent enterprise use.
Industry applications for Imagen center on automated marketing asset creation and product visualization using the same generation and customization capabilities. Automated marketing asset creation uses text-to-image generation to produce campaign visuals, banners, and variations quickly from prompts instead of manual photoshoots. Product visualization uses image editing plus subject and style tuning to show the same product in new colors, settings, or brand styles while keeping its identity consistent. Captioning and VQA support the workflow around those visuals by turning images into searchable text for catalogs, accessibility, and customer experience. For example, data flows from a product image to Imagen for captioning, and the resulting text flows to a search or storefront system.
Enterprise deployment of Imagen on Vertex AI means running image workloads on fully managed infrastructure with generative AI APIs, rather than managing servers or specialized kernels. The application sends a prompt or image request to the Imagen API on Vertex AI, Vertex AI runs inference on the foundation model, and the generated image or text returns to the application. Vertex AI Inference supports both online inference for real-time individual requests and batch inference for groups of requests returned together in one output. Vertex AI supports scalable enterprise control through regional APIs and centralized model access. Generative AI on Vertex AI can be accessed through 12 regional APIs in North America, Europe, and Asia, and regional APIs let customers control where data is stored at-rest. Models are discovered and served through Vertex AI, including the Vertex AI Model Garden for exploring Google and third-party models, so teams depend on one platform for access, deployment, and inference.
Choosing Imagen for image work depends on the type of content the workflow must produce. Google Cloud routes generative tasks by output type so each foundation model handles the work it was trained for:
| Task | Model to use | Choose when |
|---|---|---|
| Generating images | Imagen | The output must be a new image or an edited image |
| Generating code | Codey | The output must be code completion, generation, or chat |
| Generating embeddings | text-embedding | The output must be vectors for search, classification, or clustering |
| Generating text with images or video in the prompt | Gemini | The input is multimodal and the output is text |
Exam tip: Image generation and image editing in Imagen are restricted access features, while visual captioning and VQA are not listed as restricted.
Imagen (opens in a new tab) is a Google foundation model for generating and editing images, offered through the Gemini Enterprise Agent Platform. Its enterprise strengths come from pairing high-quality image creation with editing, customization, and configuration controls that teams can apply consistently. Those creation controls work together with responsible AI settings that let organizations govern what the model produces.
Enterprise image capabilities in Imagen cover both generation of new images and editing of existing images, so teams can modify work instead of starting over. Editing includes inserting objects with inpaint, removing objects with inpaint, expanding image content with outpaint, and replacing an image background. For retail and ecommerce, Imagen can also generate Virtual Try-On images. Beyond editing, Imagen offers several customization paths that let an organization shape output to its own brand or subject matter:
Responsible AI safety controls in Imagen are configurable settings that let administrators decide how the model filters content. Responsible AI and usage guidelines define what counts as acceptable use of the model. To help steer generation toward acceptable results, Imagen provides a prompt rewriter that refines user prompts, and negative prompts that let users omit content they do not want in the output. Filtering is therefore a decision the organization makes, not a single fixed behavior.
Exam tip: Imagen's safety behavior is configurable through Responsible AI safety settings — filtering is a decision the organization makes, not a fixed, unchangeable default.
Configuration parameters for production use in Imagen are the settings that control how images are produced and delivered. Aspect ratio and output resolution determine the shape and size of the generated image, and the text prompt language can be set to match the user's input. Deterministic image generation allows the same prompt to produce repeatable results, which matters when a workflow needs consistent output across runs.
Inpainting in Imagen changes content inside the boundaries of an existing image while preserving the rest of the picture, such as inserting objects, removing objects, replacing the background, or adjusting details with guided prompts. Outpainting in Imagen expands image content beyond the original edges by generating a natural continuation that extends the scene outward using the original image as visual context.
You should choose Imagen when the output must be a new image or an edited image from a text prompt or an existing image. By contrast, Codey is for code completion, generation, or chat, text-embedding models are for vectors for search, classification, or clustering, and Gemini is for text output when the input includes images or video.
Beyond text-to-image generation for creating original assets, Imagen provides prompt-guided image editing, image visual captioning that generates a text description of an image, and Visual Question Answering (VQA) that answers questions about image content. It also offers subject customization, style customization, controlled customization, instruct customization, and upscaling to increase resolution for consistent enterprise output.
Imagen provides configuration parameters including aspect ratio, output resolution, prompt language, and deterministic image generation that allows the same prompt to produce repeatable results. For safety, it provides configurable Responsible AI safety settings along with a prompt rewriter that refines user prompts and negative prompts that omit unwanted content, so filtering is a decision the organization makes.
A creative team at a retail company is preparing marketing assets for a seasonal campaign. They have high-quality photos of their merchandise and want to create multiple lifestyle advertisements by placing the merchandise into new contextual environments (such as a sunny beach, an autumn park, or a modern studio) while keeping the original product's appearance unchanged in the foreground.
Which foundation model and capability should the team use to achieve this goal?
Last updated