A marketing tool must create a new illustration from a description. Which model capability is needed?

1. Text embeddings only
2. Image generation
3. Speech recognition only
4. Entity extraction only