Question 1

What should Alex focus on for accurate Sora video generation?

❌ Ensure video under 2 minutes
❌ High-resolution monitor
❌ Length of text prompt
✅ Crafting detailed and structured text prompts

Explanation:
Sora relies heavily on prompt detail—well-structured prompts produce accurate, controlled video outputs.

---

Question 2

What should Chris include in a Sora video prompt?

❌ Font style, music
❌ Color scheme, alignment
❌ Duration, frame rate, resolution
✅ Scene context, visual details, and motion

Explanation:
Clear scene description + visual details + camera/motion cues help Sora generate high-quality videos.

---

Question 3

Which stage converts an image into a model-ready format?

❌ Data augmentation
❌ Multimodal LLM processing
❌ Image validation alone
✅ Input processing

Explanation:
Input processing handles preprocessing—resizing, normalizing, encoding—to prepare the image for the model.

---

Question 4

What combines visual features + text embeddings?

❌ Language generation component
❌ Image validator
✅ Multimodal fusion layer
❌ Visual encoder

Explanation:
The fusion layer merges image embeddings and text embeddings into one representation for captioning.

---

Question 5

Which step prepares images by encoding them?

❌ Initialize model
❌ API setup
❌ Create message
✅ Encoding images to bytes for LLM processing

Explanation:
Images must be converted into encoded tensor/byte format before the model can process them.

---

Question 6

Which component rejects invalid images?

❌ Language generation
❌ Text embedding
❌ Fusion layer
✅ Image validation step during preprocessing

Explanation:
The validation phase checks format, size, safety, and feasibility—rejecting unusable images early.

---

Question 7

Which three stages describe the multimodal captioning pipeline?

❌ Dataset labeling, caption selection
❌ Tokenization, grammar checking
❌ Editing, tagging
✅ Input processing → image validation & encoding → multimodal LLM processing

Explanation:
This is the standard workflow: prepare input → validate/encode → generate caption.

---

🧾 Summary Table