Lesson 5 of 6 · 4 min
Requesting Revisions
Use the revision dialog to give the pipeline targeted feedback on a completed output and receive a regenerated result, up to three times per video.
Requesting Revisions
Video coming soon
Lesson Notes
Each completed AI output supports up to three revision requests. A revision reruns the same workflow with your feedback appended to the base processing prompt, so the GPU worker regenerates the output with your corrections taken into account. To request a revision, click the circular-arrow revision icon beneath any video in the AI Generated Videos section. This opens the Request Revision dialog.
The revision dialog shows the current video thumbnail so you can reference it while describing what to change. It then presents two input mechanisms: Quick Suggestions and a free-text field. Quick Suggestions are preset tags you can toggle on or off: 'Wrong furniture style', 'Too dark/bright', 'Unrealistic', 'Wrong room type', 'Remove specific item', and 'Different angle'. You can select multiple tags at once. These tags are concatenated with your free-text feedback and sent as a combined prompt to the backend.
The free-text field accepts up to 200 characters (a live counter shows remaining space). Write a specific description of what should change — for example, 'The sofa looks too modern for this farmhouse-style home. Can we get something more rustic?' is more actionable than a vague 'looks wrong'. The Submit Revision button is disabled until at least one tag is selected or the text field contains text. You must provide either a tag or free text — you cannot submit an empty revision.
The dialog shows a 'Revisions remaining: X/3' counter in the bottom-left. When only one revision remains, an amber warning 'Last revision available' appears. After the third revision the revision icon on that specific video card is disabled — no more revisions can be requested for that output. Revisions per-media are tracked independently, so other outputs in the same project are unaffected.
After submission, the dialog closes, the backend creates a revision request record linked to the original job, and the pipeline worker picks it up on its next poll cycle. The project detail page resumes polling and the status badge will cycle back through Processing states until the revision completes. When done, the new output replaces the prior version in the AI Generated Videos grid.
Key Takeaways
- Each AI-generated video supports up to 3 revision requests; the count is tracked per output, not per project.
- Use Quick Suggestion tags (e.g., 'Too dark/bright', 'Wrong room type') alone or combined with free text for faster, more precise feedback.
- The free-text field caps at 200 characters — be specific and actionable rather than vague.
- The Submit Revision button requires at least one tag or non-empty text; submitting blank is blocked.
- After submission, the project detail page auto-polls and the revised output replaces the prior version when complete.