AI Image Generation Setup

AI image generation is an optional feature. When enabled, users can generate images for project covers, character portraits, and other visuals.

For how users interact with this feature, see the User Guide: AI Image Generation.

Supported Providers

Provider	Requirements
OpenAI	OpenAI API key
OpenRouter	OpenRouter API key
Fal.ai	Fal.ai API key
Stable Diffusion	Self-hosted AUTOMATIC1111 WebUI with --api flag
Cloudflare Workers AI	Cloudflare Account ID + API Token

Configuration

Image generation is set up on two admin pages: Admin → AI Providers holds the API keys, and Admin → AI Images holds the on/off switch and the model profiles users pick from.

Global Toggle

Enable Image Generation controls whether the feature is available to users. When disabled, no generation buttons appear in the UI.

Image Model Profiles

Profiles define which AI models users can choose from. Each profile wraps a provider + model combination. A profile card shows its provider and model, whether it accepts an input image, and its supported sizes; use the switch to hide a profile from users without deleting it, or the pencil and trash buttons to edit or delete it.

To create a profile:

1. Click Create Profile
2. Fill in Name (what users see) and an optional Description
3. Pick a Provider, then choose the model (see below)
4. Set the capability switches and sizes
5. Click Create Profile (or Save Changes when editing)

Which fields appear depends on the provider:

Field	Applies to	What it does
Model Type	Fal.ai	Text to Image or Image to Image. Choose it first; it filters the model list and sets Supports Image Input.
Model	OpenAI, OpenRouter, Fal.ai, Workers AI	OpenAI offers a fixed list. The others search the provider's live model list (use the refresh button if it is empty or your API key just changed).
Model ID	Stable Diffusion	Typed in by hand: the model identifier from your Stable Diffusion API.
Enabled	All	Whether users can pick the profile. Same switch as on the profile card.
Supports Image Input	All except OpenAI and OpenRouter	Lets users attach reference images. OpenAI and OpenRouter set this automatically.
Custom Resolutions	All except OpenRouter	Lets users enter any width and height instead of picking a listed size.
Aspect Ratio Only	All except OpenRouter	For models that accept an aspect ratio (such as 16:9) rather than pixel dimensions; the size list then shows ratios.
Supported Sizes / Default Size	All except OpenRouter	Sizes users can choose, as 1024x1024, and the one pre-selected. OpenRouter models use fixed aspect ratios, configured automatically.
Sort Order	All	Lower numbers appear first in the user's profile list.
Advanced: Model Config JSON	All	Provider-specific options as a JSON object, for example {"quality": "hd", "style": "vivid"}.

Provider Setup

On Admin → AI Providers, expand a provider card and add your API key to enable it.

Provider	API Key Source
OpenAI	platform.openai.com/api-keys
OpenRouter	openrouter.ai/keys
Fal.ai	fal.ai/dashboard/keys
Stable Diffusion	Self-hosted — enter your WebUI endpoint URL
Workers AI	dash.cloudflare.com — Account ID + API Token

Environment Variables

Providers can be configured via environment variables. A key or setting saved in the admin panel takes precedence.

AI_IMAGE_ENABLED=true

# OpenAI (or any OpenAI-compatible API)
AI_IMAGE_OPENAI_ENABLED=true
OPENAI_API_KEY=sk-...
# OPENAI_API_BASE=https://your-compatible-endpoint/v1

# OpenRouter
AI_IMAGE_OPENROUTER_ENABLED=true
OPENROUTER_API_KEY=sk-or-...

# Fal.ai
AI_IMAGE_FALAI_ENABLED=true
FALAI_API_KEY=fal-...

# Stable Diffusion
AI_IMAGE_SD_ENABLED=true
SD_API_ENDPOINT=http://localhost:7860
# SD_API_KEY=...   # only if your endpoint requires one

# Cloudflare Workers AI
AI_IMAGE_WORKERSAI_ENABLED=true
WORKERSAI_ACCOUNT_ID=your-cloudflare-account-id
WORKERSAI_API_TOKEN=your-workers-ai-api-token

These keys are shared with the AI text features (linting, Auto-Review, prompt optimization), which read OPENAI_API_KEY, OPENROUTER_API_KEY and ANTHROPIC_API_KEY from the same place. The environment does not create image profiles; add those in the admin panel.

note

Image generation is also gated by the AI Kill Switch, which is on by default. Turn it off before expecting any AI feature to appear.

Image audits

Admin → Image Audits (hidden while the AI kill switch is on) lists every image generation that completed or was moderated, newest first. Each record keeps:

• the date and the user who made the request,
• the profile used and the prompt text,
• the credits charged (the profile's credit cost),
• the status: success, or moderated when the provider returned a text refusal instead of an image,
• links to the generated image where one was stored, plus the reference images and any moderation message in the underlying record.

The page header shows totals for Total Requests, Total Credits Used, Successful and Moderated. Search the prompt text with Search prompts, filter by Status, and page through the results (10 to 100 per page); the view generated image button opens the output. Requests that fail before the provider responds (for example a network error) are not recorded. Reference-image URLs and a moderation message are also stored with each record, and are available through the audits API (/api/v1/admin/image-audits).

Browser and Cloud Sync profiles

AI image generation requires a server connection and is not available with a Browser or Cloud Sync profile.

Cost Considerations

Costs vary by provider. Check your provider's pricing. Self-hosted Stable Diffusion has no per-image cost.

Troubleshooting

• Provider not available: Verify API key is correct and has credits
• Generation failures: Content policy violation, rate limits, or network issues
• Stable Diffusion not connecting: Ensure WebUI is running with --api flag

API keys are stored encrypted and never displayed after saving.

Related

• User Guide: AI Image Generation — How users interact with this feature
• AI Kill Switch — Emergency disable for AI features