Prepare local AI processing

Identify local AI nodes

File Processing nodes such as background removal, inpainting, upscaling, and video depth use their own local models. Their configuration is separate from cloud chat providers, API keys, and credits.

First-time setup

After adding the required nodes, explicitly choose Download and install in the runtime prompt. Runtime packages, components, and model weights may need network access and disk space. Wait for readiness before running a sample. Check connectivity and space if downloading fails, then use Retry.

Check output

Inspect edges and transparency for background removal, masked regions for inpainting, dimensions and detail for upscaling, and frames and duration for depth output. Test a small file before a batch. Performance depends on hardware, model, and input size.

Successful installation does not prove every workflow’s output is correct. Keep originals and record nodes and settings. Use the feedback guide for reproducible issues.