Generation Failed

Every node in Ozu moves through a predictable set of states as it works. Most of the time it runs through them silently and you never look. When something stops, the node’s state tells you what happened and what to do next. This page maps the states, explains the difference between a failed generation and an error, and covers how to get unstuck.

​

The pipeline states

And two states that mean something stopped:

• Failed — the generation was attempted and didn’t succeed. Usually transient.
• Error — the generation couldn’t run because of a structural problem in the inputs or dependencies.

Same visual distinction on the node. Different meanings. Different fixes.

​

Failed versus Error

​

Failed

A failed state means the generation started but didn’t return a usable output. The common causes are transient and not your fault:

• Network hiccup — a request timed out mid-flight
• Model rate limit — the AI provider throttled the request
• Provider issue — the provider’s service had a temporary problem
• Safety block — content flagged by the provider (see Safety Blocks)

For failed states, retry the node. Most transient failures resolve on a second attempt. Credits from the failed attempt were released — you won’t be double-charged.

​

Error

An error state means the node couldn’t even attempt the generation. Something upstream is wrong:

• Missing dependency — an input asset isn’t present or hasn’t been approved
• Invalid input — a required field is blank or malformed
• Asset not approved at the Director Review Board — downstream frames can’t run until every asset has passed DRB
• Blocked by Storyboard or Timeline state — the Storyboard Node needs approved assets; the Timeline Node needs rendered frames

For errors, fix the underlying problem before retrying. The node will tell you what’s missing — usually in the details panel or as a tooltip on the status icon.

​

What happens to your credits

Credits move through three stages:

1. Held when a generation starts — reserved against your balance, not spent
2. Settled (debited) on a successful complete
3. Released (returned) on failure, error, or safety block

You are never charged for a generation that didn’t produce a usable output. If a node fails, its held credits come back to your balance. Your balance indicator in the user menu updates within a few seconds.

​

How to retry

Every node has a retry action. The exact control depends on the node, but the pattern is consistent:

• Single-frame retry — regenerate one Frame from the Storyboard Node. Use this for a specific shot that came out wrong.
• Asset reshot — regenerate the underlying Actor, Prop, or Set from its own node. This rebuilds every downstream Frame that uses that asset.
• Node retry — re-run the node from its status indicator. The most common case.

If a node keeps failing, treat it as a signal. Two failures in a row can be transient. Five in a row means something about the prompt, the input, or the provider is systematically wrong — adjust before retrying again.

​

Common scenarios

A single frame keeps failing. Try a different prompt wording. If it still fails, the cause is likely a safety block — check the node details for a block reason. See Safety Blocks. A whole scene errored out. Walk back through the Flow canvas. Something upstream — usually an unapproved asset — is blocking the run. Approve it at the DRB, then retry. The Timeline Node won’t render. It needs every referenced frame to have completed. Find any frame still in failed state, retry it, then re-run the Timeline. An asset reshot succeeded but old frames still look wrong. Reshots rebuild downstream frames, but the rebuild is explicit — you’ll see the downstream Storyboard Node return to pending while it regenerates the affected frames. Give it a minute, then check. Every generation is failing across the project. Likely a provider-side issue. Wait fifteen minutes and try again. If it persists, contact support.

​

When to contact support

Most failures resolve with a retry or a small prompt adjustment. Contact support if:

• Every generation across your project has been failing for more than thirty minutes
• Credits were debited from your balance but the node is marked failed
• A node’s error message doesn’t match any of the patterns above
• You see repeated errors with no clear upstream cause

To reach support, use the contact channel listed in the Ozu app menu or on the main Ozu website. Include your project name, the node in question, and the approximate time the issue started.

​

Related

​

FAQ

Does a failed generation cost credits? No. Credits are held at the start of a generation and released if it doesn’t complete successfully. Your balance is not charged for a failed attempt. What’s the difference between failed and error? Failed means the generation ran and didn’t succeed — usually transient, usually fixed by retrying. Error means the node couldn’t start because an input or dependency is missing — fix the upstream problem first. How many times should I retry before giving up? Two or three retries is reasonable for a transient failure. If it fails five times in a row, the cause is systemic — adjust the prompt, check for a safety block, or contact support. Can I cancel a running generation? Most nodes let you cancel while running — the node returns to idle and any held credits are released. Nodes near the end of a long generation may complete before the cancel registers. If my internet drops, does my generation survive? The generation itself runs on our side, not yours. You can close the tab, lose connection, and come back — the node keeps running and its state updates when you’re back online. I see “credits held” but the node looks idle. Is that a bug? The hold is momentary as the node transitions from pending into running. If held credits persist on an idle node for more than a minute, contact support with the project and node details.