1.8 Generative Media and Models as a Service: Gemini, Imagen and Veo
Key Takeaways
Gemini handles multimodal understanding and text generation, Imagen generates and edits images, and Veo generates video — each with distinct tuning, latency, and billing characteristics.
Models as a service in Model Garden lets partner and open models be called through the same Agent Platform API surface, IAM model, and billing account as first-party models.
Imagen supports subject and style customization from a small set of reference images, which is the correct answer for brand-consistent image generation.
Video and image generation are long-running operations billed per generated asset or per second of output, so they belong in asynchronous workflows rather than synchronous request paths.
SynthID watermarking and safety filtering apply to generated media, and are the controls a provenance or misuse question is pointing at.
1.8 Generative Media and Models as a Service: Gemini, Imagen and Veo
Blueprint reference: Section 1.2, "Building solutions and tuning models for specific use cases (e.g., Gemini, Imagen, Veo, and models as a service in Model Garden)."
The exam does not ask you to write prompts. It asks you to pick the right generative family for a described business outcome and then to architect around that family's operational shape. Three of those shapes are quite different from each other.
The Three First-Party Families
| Family | Input | Output | Shape of the call | Typical billing |
|---|---|---|---|---|
| Gemini | Text, images, audio, video, PDFs | Text, structured output, tool calls | Synchronous, streaming supported | Per input/output token |
| Imagen | Text prompt, optional reference or mask images | Images; also edit, upscale, inpaint, outpaint | Short asynchronous or synchronous | Per generated image |
| Veo | Text prompt, optional conditioning image | Video clips | Long-running operation | Per second of generated video |
Gemini is the general-purpose reasoning and understanding model, and the only one of the three that natively consumes multimodal input to produce text. When a scenario says "extract fields from a scanned invoice," "answer questions about this video," or "classify this support ticket," the family is Gemini even though the input is an image or video.
Imagen produces and manipulates images. Its capabilities beyond straight text-to-image are the ones that show up in scenarios: inpainting (replace a masked region), outpainting (extend beyond the original frame), upscaling, and subject or style customization from a handful of reference images. That last one is the answer whenever the requirement is brand consistency — a specific product, mascot, or house visual style reproduced across many generated assets.
Veo generates video and behaves like a batch job. A request returns an operation handle; the client polls or receives a completion signal later. Any architecture that blocks an HTTP request on video generation is wrong.
Models as a Service
Models as a service (MaaS) means calling a partner or open model through the Agent Platform's own API surface rather than through the vendor's. The practical consequences are what the exam tests:
- One IAM model. Access is granted with Google Cloud IAM rather than a separately-managed vendor API key.
- One bill. Usage appears on the Google Cloud invoice, inside existing budgets and committed-spend agreements.
- One network path. Traffic stays within Google Cloud's boundary rather than egressing to a third-party endpoint.
- No infrastructure. Unlike deploying an open checkpoint yourself, MaaS models are managed endpoints with no accelerators to size or pay for while idle.
When a scenario complains about managing separate vendor keys, unpredictable third-party invoices, or security review of external endpoints, MaaS is the intended answer.
Tuning Across the Families
Customization means different things per family, and conflating them is a common error.
- Gemini — supervised tuning on prompt/response pairs, and adapter-based tuning that produces a tuned endpoint. Teaches behaviour, format, and domain style.
- Imagen — subject customization and style customization from reference images. Teaches what a thing looks like, not how to reason.
- Veo — customization is primarily prompt-level and conditioning-image-level; there is no general-purpose supervised tuning workflow comparable to Gemini's.
- MaaS partner models — tuning availability depends entirely on the partner; assume none unless the model card says otherwise.
Architecting Around Long-Running Generation
Media generation belongs in an asynchronous pattern. The reference shape:
Client request
└─> API service writes a job row (status = QUEUED) and publishes to Pub/Sub
└─> Worker (Cloud Run) calls Imagen or Veo, receives an operation handle
└─> Worker polls or awaits completion, writes the asset to Cloud Storage
└─> Job row updated to READY; client notified or polls status
Three reasons this matters on the exam. First, timeouts: a synchronous HTTP request cannot wait out a video render. Second, cost control: a queue lets you apply per-user quotas and deduplicate identical requests before spending on generation. Third, retries: generation failures are common enough that idempotent job records beat naive client retries, which otherwise pay twice for the same asset.
Provenance and Safety on Generated Media
Two controls appear in governance-flavoured questions:
- SynthID embeds an imperceptible watermark into generated images and video, allowing later verification that an asset was AI-generated. This is the answer to "how do we prove which assets our systems produced" and to internal misuse-detection requirements.
- Safety filters screen both prompts and generated output. Configurable thresholds control blocking behaviour, and requests can be rejected at the prompt stage rather than after paying for generation.
For text-based generative applications, Model Armor adds prompt-injection and jailbreak screening on top of content safety filtering; that control is covered with the rest of Section 6.
Choosing the Family: A Fast Decision Table
| Requirement | Family |
|---|---|
| Understand or answer questions about an image, PDF, audio clip, or video | Gemini |
| Produce marketing images that match a brand's existing product photography | Imagen with subject/style customization |
| Remove an object from an existing photo | Imagen inpainting |
| Extend a photo to a wider aspect ratio for a banner | Imagen outpainting |
| Generate a short promotional video clip from a script | Veo |
| Call a third-party frontier model on the Google Cloud bill with IAM-based access | MaaS in Model Garden |
| Run a specific open checkpoint with weights under your control inside your VPC | Self-deployed open model (not MaaS) |
Exam Traps
- Using Imagen to "understand" an image. Understanding is Gemini; Imagen generates and edits.
- Synchronous video generation. Veo is a long-running operation.
- Assuming partner models can be tuned. Check the model card.
- Confusing MaaS with self-deployment. MaaS has no endpoint to size; self-deployment does.
- Forgetting per-asset billing. Image and video generation is not billed per token, so token-based cost controls do not apply.
A marketing platform must generate 15-second promotional video clips on demand from user-supplied scripts. The current prototype calls the generation API directly inside the web request handler and users report frequent gateway timeouts. What is the correct architecture?
Raise the load balancer and application timeouts to accommodate the generation duration and add client-side retries.
Accept the request, enqueue a job, generate with Veo in an asynchronous worker that writes the finished clip to Cloud Storage, and notify the client when the job record turns ready.
Pre-generate a library of clips nightly with a batch job and serve the closest match to each user request.
Switch from Veo to Imagen and animate the resulting images client-side to reduce generation time below the timeout.
A retailer needs thousands of lifestyle images that feature one specific sneaker model, matching the exact colourway and silhouette of their existing product photography. Text prompts alone produce plausible but inconsistent shoes. Which capability addresses this?
Gemini multimodal generation, supplying a product photograph with each prompt as grounding context.
Supervised tuning of a Gemini model on captioned product photographs.
Imagen subject customization using a small set of reference images of the specific sneaker.
Imagen outpainting applied to the existing product photographs to widen them into lifestyle scenes.
A security review objects to a proposal that calls a third-party frontier model through the vendor's own public API, citing separate API key management, an invoice outside cloud budgets, and traffic leaving Google Cloud. Which approach resolves all three objections without self-hosting?
Deploy the partner model's open-weight checkpoint to an Agent Platform Inference endpoint inside the VPC.
Replace the partner model with a first-party Gemini model of equivalent capability.
Store the vendor API key in Secret Manager and route calls through a Cloud Run proxy inside the VPC.
Consume the same partner model as a service through Model Garden, so access is IAM-governed, usage lands on the Google Cloud invoice, and traffic stays within Google Cloud.
A media company must be able to demonstrate, months later, which images circulating internally were produced by its own generative pipeline. Which control provides this?
SynthID watermarking, which embeds an imperceptible marker into generated images that can be verified afterwards.
Cloud Audit Logs on the generation endpoint, correlated with the request identifiers stored alongside each asset.
Customer-managed encryption keys on the Cloud Storage bucket that holds the generated assets.
Configuring safety filters at a strict threshold so that only compliant images are generated.
Sections you finish are checked off in the contents.