The Providers tab lists every image provider the plugin can call, split into free and paid sections. Each row has a masked key field with Save Key, Test Key and Clear buttons.
Providers that need no key:
- Pollinations.ai: the default. Works with no key at all. An optional free key from its own signup page moves you to a higher quality tier.
- AI Horde: crowdsourced GPUs, free anonymously. A free signup key is optional and gives you higher queue priority.
Free tier providers that do need a key:
- Cloudflare Workers AI: free tier, key format is your account ID and API token joined with a colon.
- Hugging Face: free monthly credits. The token needs the Inference Providers permission.
- Together AI: a free FLUX Schnell endpoint with a free key.
Paid providers, billed by the provider directly:
- Google, Gemini image models, generating on Nano Banana 2 by default. Also one of the two character lock routes, and the free API key from Google AI Studio also powers the text features.
- FLUX via fal.ai. The other character lock route, and the same key drives the premium upscaler.
- OpenAI GPT Image, on gpt-image-2. Also Stability AI, Ideogram, Recraft, Leonardo AI and xAI Grok Image.
How are my keys stored?
Keys are encrypted at rest with AES-256-GCM using a site specific salt, falling back to AES-256-CBC with an HMAC when GCM is not available on the server. They are never written to log files, and a key is only ever sent to the provider it belongs to.
The Test Key button checks a key before a whole book depends on it. For OpenAI, Hugging Face and Together AI it makes a live call to the provider and shows the answer on the row. For the other providers it confirms that a key is saved, and a key the provider does not accept shows up on the first page that fails, with the provider's own error message on that page's card. Use Test Key every time you paste a new key.
Older models upgrade automatically:
Providers retire image models from time to time. If a saved setting or an older book still names a retired Google or OpenAI image model, the plugin uses the current one instead, a current Gemini image model (Nano Banana 2 by default) for Google and gpt-image-2 for OpenAI, so your existing books keep generating without you changing anything.