Providers
Image generation providers available in Content Studio — vendor, model IDs, API key source, and capability summary.
Content Studio routes asset generation through vendor APIs for image creation. The platform selects the active provider based on workspace configuration and falls back automatically when the primary provider is unavailable.
Image providers
| Provider | Vendor | Model IDs | Default | Sensitivity |
|---|---|---|---|---|
| Gemini Imagen 3 | Google (Vertex AI) | imagen-3.0-generate-001 | Primary | Standard |
| OpenAI | OpenAI | gpt-image-1, dall-e-3 | Fallback | Standard |
Gemini Imagen 3
Vendor: Google Cloud — Vertex AI
Model: imagen-3.0-generate-001
API key source: Authenticated via Application Default Credentials (ADC) on the server side. No API key is required from workspace owners. The platform service account holds the necessary IAM role.
Capabilities:
- Slide and infographic image generation at five aspect ratios:
1:1,3:4,4:3,9:16,16:9 - Deterministic seeding — the same prompt and seed value always produce the same image, enabling narration edits without regenerating artwork
- Style preset injection for visual consistency across a run
- Up to 3 attempts with exponential back-off (2 s, 4 s) on transient quota and service errors
Sensitivity: Prompts are screened by Google Responsible AI policies. Requests that trigger safety filters return an error and the run marks that scene as failed. No image data is stored after delivery.
OpenAI
Vendor: OpenAI
Models: gpt-image-1 (primary), dall-e-3 (fallback within the OpenAI provider)
API key source: The platform holds an OpenAI API key provisioned at the workspace level. To use your own OpenAI key instead of the platform's, see Bring Your Own Key.
Capabilities:
- Slide and infographic image generation
- Aspect ratios:
1:1,3:4,4:3,9:16,16:9 - Automatic internal fallback from
gpt-image-1todall-e-3whengpt-image-1is not available on the account tier - Three-attempt exponential back-off on rate-limit and server errors
Sensitivity: OpenAI content policy applies. Requests rejected by policy return an error; no image bytes are stored.
Provider selection and fallback
Content Studio uses Gemini Imagen 3 as the primary image provider. When Imagen 3 returns an error after exhausting its retry budget, the request is automatically forwarded to OpenAI exactly once.
The fallback fires on provider failures only — safety-policy rejections from either provider propagate as errors without triggering the other provider.
| Condition | Outcome |
|---|---|
| Imagen 3 succeeds | PNG bytes returned from Imagen 3 |
| Imagen 3 fails (quota, 403, timeout) | Up to 3 attempts internally, then forwarded to OpenAI once |
| OpenAI also fails | Run scene marked failed; credits are refunded |
| Safety rejection from either provider | Error returned immediately; no cross-provider retry |
Sensitivity classification
Content Studio image generation is subject to the platform's content safety policy. The same flags that govern ingestion apply:
| Flag | What it detects | Outcome for image generation |
|---|---|---|
v | Violence | Prompt rejected before generation |
x | Sexually explicit | Prompt rejected before generation |
n | Nudity | Allowed with content warning |
Prompts that contain extracted workspace node data inherit the sensitivity flags of the contributing nodes. Nodes marked confidential: true are excluded from visual prompts before they reach the image provider.
Content Studio
Generate video, audio, and infographic content from workspace knowledge graph data — configuration reference for providers, model selection, and API keys.
Model Selection
How to choose which model powers a Content Studio generation — the per-generation model selector in the chat composer.