# Multimodal and Generative Capabilities

Modern AI generates and understands text, images, audio, and video together. This sub-article expands [Section 2 of Recent Topics](../recent.md).

## Text-to-Image and Video
- **DALL·E**, **Stable Diffusion**, **Midjourney** generate images from text.
- **Sora**, **Veo** generate video; diffusion models are central.

## Voice and Audio
- **ElevenLabs** and similar tools clone voices.
- Speech synthesis becomes hard to distinguish from human speech.

## Multimodal Understanding
- Models jointly process text, image, audio, and video for search, accessibility, robotics.
- Vision-language-action models connect perception to physical action.

## Long-Context Models
- Million-token contexts enable whole-book reasoning and large-document analysis.

## Why It Matters
Multimodal generation expands AI's reach into creativity, communication, and physical action.

## Sources
- OpenAI. *DALL·E and Sora documentation.*
- Stability AI. *Stable Diffusion.*

*Back to [Recent Topics](../recent.md).*