Draft generation, AI recommendations, and security fixes in ClassifAI 3.4.0

ClassifAI 3.4.0 makes AI-assisted publishing more practical for day-to-day editorial work, led by a new content generation feature that can turn a short summary into an initial draft. This release also adds an AI-powered Recommended Content block, expands WooCommerce support, updates several AI models, and includes security and reliability fixes.

Important upgrade note: ClassifAI now requires WordPress 6.7 or newer, and this release is tested up to WordPress 6.8.

Feature highlights

• Draft content generation — Start with a summary of what you want to publish and generate a draft using supported providers including OpenAI, Azure OpenAI, and Ollama.
• Recommended Content block — Add a block that suggests related content using OpenAI embeddings, giving readers a more useful path to additional posts or pages.
• Recommended Content defaults — Configure global default template settings so recommended content output is easier to keep consistent across your site.
• New OpenAI image model — Generate images with support for OpenAI’s newer gpt-image-1 model.
• WooCommerce writing support — Generate product-focused titles and excerpts using details such as product type, SKU, categories, attributes, pricing, stock status, and descriptions.
• Manual image processing mode — Choose whether image processing features run automatically on upload or only when someone manually requests them.
• New transcription models — Select OpenAI’s GPT-4o mini Transcribe or GPT-4o Transcribe models for audio transcription workflows.
• New text-to-speech model — Use OpenAI’s GPT-4o mini TTS model, which is now the default for the Text to Speech feature.

Other improvements and fixes

• Google AI model update — The Gemini provider now uses Gemini 2.5 Flash instead of the discontinued Gemini Pro model.
• Editor polish — The Generate Titles modal has an updated layout, registration messaging has been refined, and admin notices are less intrusive.
• Settings access — Feature settings remain accessible even when a feature is turned off.
• Reliability fixes — The release addresses a fatal error when $post was null, PHP notices and warnings, Text-to-Speech post type availability, unnecessary embeddings notices, Gutenberg deprecation notices, and stale content being sent to title and excerpt generation.
• Security updates — Several dependencies were updated for security, including tar-fs, react-router, http-proxy-middleware, and axios.

For developers

This release includes a few provider naming changes, new filters, and additional documentation for teams extending or auditing ClassifAI integrations.

• OpenAI image provider rename — The DallE provider class is now Images, and related DALLE references have been updated.
• OpenAI speech provider rename — The Whisper provider is now SpeechToText, so custom extensions that subclassed Whisper should update their references.
• Google AI model filter — The new classifai_googleai_gemini_api_model filter makes it easier to override the Gemini model.
• Speech to Text filters — The new classifai_openai_speech_to_text_model and classifai_openai_speech_to_text_api_url filters allow model and endpoint customization.
• Text to Speech filters — The new classifai_openai_text_to_speech_model and classifai_openai_text_to_speech_api_url filters allow model and endpoint customization.
• Developer documentation — New docs cover supported providers, which features each provider supports, how data is used, and initial data flow diagrams.

See the full release notes on GitHub.

Learn more and download ClassifAI.