T2I-Adapter for Conditioned Synthesis Quiz

Answers and explanations appear as you take the quiz.

1. What is the main advantage of T2I-Adapter over ControlNet?
   • It is much lighter, with far fewer trainable parameters
   • It generates audio as well as images
   • It requires no condition input at all
   • It only works at 256x256 resolution
2. Which of these is a valid condition input for T2I-Adapter?
   • A spreadsheet of weights
   • A depth map or pose skeleton
   • An audio waveform
   • A list of hyperparameters
3. Why is T2I-Adapter computationally efficient at inference time?
   • It skips the text prompt entirely
   • It uses no neural network
   • It computes condition features once instead of re-processing at every denoising step
   • It only runs on the final pixel image
4. What happens to the base diffusion model when you add a T2I-Adapter?
   • It is fully retrained
   • It is deleted and replaced
   • Its text encoder is removed
   • It stays frozen while only the adapter is trained
5. Can multiple T2I-Adapters be used together?
   • Yes, for example combining pose and depth to compose richer scenes
   • No, only one condition is ever allowed
   • Only if you retrain the base model
   • Only with audio adapters
6. How does the adapter's output reach the diffusion model?
   • It replaces the VAE entirely
   • Its multi-scale features are added to the frozen U-Net encoder's matching resolution levels
   • It is appended to the text prompt as words
   • It modifies the random noise seed only
7. In a T2I-Adapter workflow, what does the text prompt mainly control while the condition map supplies structure?
   • The number of U-Net layers that remain frozen
   • The adapter's learned parameter count
   • The semantic content and style of the generated image
   • The coordinates used to resize the condition image
8. Approximately how many parameters does the guide give for a typical T2I-Adapter?
   • About 7.7 thousand parameters
   • About 770 million parameters
   • About 77 billion parameters
   • About 77 million parameters