A team needs to answer questions about a supplied photograph. What model capability must it select?

1. Embedding generation only with no answering component
2. Speech synthesis only
3. Support for image input and language output
4. Text-only input with no image-processing path