Mix and match AI providers per feature on the same site — pay-as-you-go for one feature, on-prem for another, premium for a third.
How it works
ClassifAI is unusual in the WordPress AI plugin landscape because every feature is configured independently: there is no single “AI provider” setting at the plugin level, only a provider dropdown on each feature’s own admin page. That means a site can drive title and excerpt generation through OpenAI ChatGPT for the best quality on short prompts, run image generation through Together AI for cost reasons, run smart cropping through Azure AI Vision because that is the only supported backend, run text-to-speech through Amazon Polly for predictable per-character pricing, run audio transcription through ElevenLabs for accent handling, and run all the embeddings features (classification, Smart 404, term cleanup) through a locally hosted Ollama instance to keep article-level vector data on infrastructure they own.
What stays consistent across providers
- Role and user gating settings.
- The programmatic-credentials filter pipeline.
- Per-feature usage tracking, so administrators can see real costs per provider per feature.
Supported providers
- OpenAI
- Azure AI (OpenAI Service, Vision, Speech)
- Google Gemini
- xAI Grok
- Ollama (locally hosted)
- IBM Watson
- Amazon Polly
- ElevenLabs
- Together AI
- Stable Diffusion via AUTOMATIC1111 (locally hosted)
Why it matters
Together with the locally hosted LLM path, multi-provider flexibility is what lets ClassifAI fit into procurement and compliance regimes — from a one-person blog on shared hosting to a regulated enterprise newsroom — without forking the plugin.