Symptom

  • As soon as you run the workflow, or partway through sampling, a red error screen appears and processing stops.

  • The terminal shows torch.cuda.OutOfMemoryError along with a message about VRAM usage.

When it happens

  • When running heavy models such as Flux or video models at a high resolution or with a large batch size.

Cause

  • The combination of model + image size + batch size is too large for the GPU's VRAM capacity.

Solution

  • Set batch size to 1, and lower resolution to the model's recommended value (around 512-768px for SD1.5, around 1024px for SDXL).
  • If that still doesn't work, use a lighter model or consider a quantized model (gguf / nf4, etc.).
  • If even that doesn't help, just accept it and upgrade your GPU.