
Cloudinary's AI preview duration participates in choosing the content. Changing that duration can select different video segments; it is not documented as merely trimming the end of the previous preview. If an approved moment must remain, distinguish shortening an existing rendition from requesting a newly analyzed summary.
Key Takeaways
- Preview length is an input to AI content selection.
- Segment-count defaults also vary with requested duration.
- A shorter summary needs its own editorial review.
Decide which kind of revision the client requested
Keep the approved preview and the original film as separate records. Ask whether the task is “make a new shorter summary” or “keep these approved moments but cut this particular ending.” Those requests can call for different workflows.
The Cloudinary preview guide explicitly says that different requested durations may result in different selected segments. It also describes controls for total duration, maximum segment count and minimum segment duration.
A hypothetical wedding teaser might have an approved glance between the couple. Requesting a shorter AI summary does not promise that this particular glance will remain, even if it was prominent in the longer preview.
Record all content-selection inputs
Keep the source revision, preview duration and any explicit max_seg and min_seg_dur values with the candidate. The official defaults change across duration bands, so modifying length while leaving other controls unspecified can also change the default segment arrangement.
For example, the captured guide lists one segment for previews up to five seconds and different multi-segment defaults for longer previews. Those are documented settings, not measured selection results or a rule about which scene deserves inclusion.
Do not compare candidates from different film revisions and attribute every changed scene to duration. Hold the source constant when evaluating that input.
Use a controlled comparison to understand the change
Prepare a non-sensitive candidate with the intended new duration through the supported generation workflow. Compare recognizable scene boundaries, subject identities and narrative order with the retained approved preview.
Write down any moment that must remain and check it explicitly. A pleasant-looking summary can still omit a product reveal, qualification or person that the brief requires.
If the task instead requires a fixed excerpt, use an appropriate deliberate trimming or editing workflow. The preview-position guide illustrates a different process in which a chosen source position defines the sample. Do not describe that control as part of e_preview.
Approve the new summary as a new artifact
Review beginning, ending, important actions and sound in the actual output. The export-range guide helps distinguish an intentionally shortened film from an unexpectedly missing ending.
Keep an approved-version record so a team does not replace the accepted scene selection simply because another candidate is shorter. Accept a newly generated summary only after its content is suitable for the brief. This procedure reports no generation result and promises neither deterministic selection nor retention of a particular moment.
Related tasks
- How to Choose a HandBrake Preview Position and Duration
- The timeline is complete but the export is only ten seconds: check the export range
- How Do I Keep the Approved AI Image or Video from Getting Lost Among Revisions?
Sources and scope
- Cloudinary: Video previews and posters: e_preview analyzes the whole video; eager or explicit generation and notification are recommended for production, while direct experimentation can initially return 423. Requested duration affects selected content. It is not supported as an incoming transformation.
Documentation-based guidance from the cited official pages captured on October 25, 2026. Scenarios and acceptance procedures are proposed, not executed tests. No account operation, measured demand, AI-citation frequency, processing benchmark or search-indexing outcome is claimed.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: AI-generated conceptual editorial illustration; not a real application screenshot, observed output or before-and-after test.