The result does not match the current project

A visible result is a saved output from a particular execution. First find out which inputs produced it and whether those inputs still match the project.

Symptom	Inspect	Next step
Old image remains after a prompt edit	Freshness indication and current run plan	Decide whether to keep it for comparison or accept the cost of updating it.
Repeated run returned the same image	Cache/reuse information and unchanged inputs	Reuse is expected; deliberate regeneration is a separate potentially paid action.
Original script differs from a scene	Current scene, rewrite records and old script artifact	Use the current scene output for downstream work; keep the original as history.
Scene address has disappeared	Fresh scene_in/scene_out addresses	A rewritten/regenerated script can have different scene identities; inspect the current one.
No provenance events	read_run_history as well as trace	Some operations do not record provenance; empty trace alone does not mean “never ran.”
Critic returned no findings	Its check status	failed or unknown does not establish a successful review.
Node result missing after successful generation	Run save state, job and materials	Follow the attachment or not_saved recovery before requesting new content.
A long report looks incomplete	Explicit truncation/omission flags	Read the full content on the canvas; do not infer missing text.

When a reference was delivered but the image or clip still misses the brief, that is a generation-quality issue rather than a missing-input issue. Compare the actual result with the request and revise one relevant cause at a time. A supported input or a successful past example does not guarantee compliance on this render.

For agent work, inspect the project, report what is known and follow run recovery if the outcome is uncertain. Read instructions and results for how text moves between nodes.

Last checked: 14 October 2026. Cache reuse, project freshness and visual quality answer different questions.