Validation

Prompt outputs failed validation

Prompt outputs failed validation

ComfyUI refused to queue the prompt because a node failed pre-run validation — an unconnected required input, an out-of-range value, or a missing file.

What it means

Before running, ComfyUI validates every node in the graph: required inputs must be connected, values must be within allowed ranges, and referenced files (checkpoints, LoRAs, VAEs) must exist. If any check fails, the whole prompt is rejected and this message appears, usually naming the offending node and field.

It is a pre-flight check, so nothing actually ran — the fix is to satisfy the validation the message points to rather than to change sampler settings.

Common causes

  • A required input on a node is left unconnected.
  • A value is outside the node’s allowed range.
  • A referenced model/file is not present in the dropdown (renamed or missing).
  • A type mismatch between two connected sockets.

How to fix it

  1. 1

    Read which node and field failed

    The validation message names the node and the specific input that failed. Start there rather than guessing.

  2. 2

    Connect required inputs

    Ensure every required (non-optional) input on the flagged node has an incoming connection of the correct type.

  3. 3

    Refresh node definitions

    If a dropdown value looks empty or stale, refresh the browser (or press R) so ComfyUI reloads model lists and node definitions, then re-select the value.

  4. 4

    Verify referenced files exist

    Confirm any checkpoint/LoRA/VAE the node references is actually present in the models folder and selectable in the dropdown.

A workflow captured with its exact model references intact validates cleanly when reopened, instead of failing because a file was renamed or the dropdown lost its selection.

Frequently asked questions

Frequently Asked Questions

Quick answers to the most common questions.

Related errors

Make every ComfyUI run reproducible

Numonic captures the full workflow behind each ComfyUI generation — models, samplers, seeds, prompts, and custom nodes — so a setup that finally works is one you can find and rebuild, not re-debug.