What it does

org-ai is a minor mode that adds #+begin_ai...#+end_ai blocks to org-mode. You type a prompt, hit C-c C-c, and ChatGPT (or a local LLM via oobabooga) responds inline. Add :image and you get DALL-E or Stable Diffusion output stored in ~/org/org-ai-images/. It also handles speech input via Whisper and text-to-speech via espeak—because apparently Emacs users want to talk to their editor now.

The interesting bit

The clever part is inheritance: set system prompts, temperature, and model as org properties on a heading, and every AI block underneath picks them up. It’s org-mode’s outline structure repurposed as prompt engineering infrastructure.

Key highlights

• Supports OpenAI, Azure, Perplexity, Anthropic/Claude, and local models via text-generation-webui
• Image generation with DALL-E 3 or Stable Diffusion, including img2img via :image-ref
• Speech input (Whisper/Siri dictation) and output (espeak/greader)
• Global commands like org-ai-on-project for prompting across multiple files outside org-mode
• Syntax highlighting inside AI blocks by appending a major mode: #+begin_ai markdown

Caveats

• Speech setup is platform-specific and involves multiple external tools (Whisper, espeak, portaudio); the README warns macOS and Windows steps differ
• Stable Diffusion requires running AUTOMATIC1111’s webui separately; local LLMs need conda environments and manual API server startup
• The “free tier gets you pretty far” note about OpenAI is the author’s impression, not a guarantee

Verdict

Worth a look if you already live in org-mode and want AI woven into your note-taking workflow, not a separate app. Skip it if you want something that works out of the box without API keys, local servers, and speech toolchain wrangling.