ComfyUI i quiz dotyczący przepływów pracy z obrazami opartymi na węzłach

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. Which node runs the denoising loop in a basic ComfyUI graph?
   • VAE Decode
   • Save Image
   • Empty Latent Image
   • KSampler
2. What three components does Load Checkpoint output?
   • Seed, steps and CFG
   • The denoising model, the CLIP text encoder and the VAE
   • Prompt, image and mask
   • GPU, CPU and RAM
3. How can you recover the workflow that made a ComfyUI image?
   • Drag the PNG into ComfyUI, because the workflow is embedded in its metadata
   • Reverse the image through a VAE
   • Search for the seed online
   • It cannot be recovered
4. Why does changing only the prompt not reload the model?
   • Models are streamed from the cloud
   • The prompt is never used
   • ComfyUI caches node results and re-runs only nodes whose inputs changed
   • ComfyUI keeps a copy of every image
5. What is the main security risk of custom nodes?
   • They slow down the interface
   • They are Python code that runs with your user account's permissions
   • They change image resolution
   • They require an internet connection
6. Which model file format is safer to load?
   • .ckpt (pickle-based)
   • .exe
   • .zip
   • .safetensors
7. Which format do you POST to ComfyUI's /prompt endpoint for automation?
   • The API prompt format, which lists nodes by ID with class type and inputs
   • The UI workflow layout file
   • A PNG file
   • A CSV of seeds
8. Which hardware does ComfyUI support best?
   • CPU-only systems
   • AMD GPUs with no extra setup
   • NVIDIA GPUs
   • Intel integrated graphics