Try a different first sentence.

Keep the promise within the video.

For a notebook-planning video, three possible openings are “Which task needs your attention?”, “Start with one clear task.” and “Leave room for a useful next step.” Each points to the same modest action. None promises that a planning habit will double productivity or guarantee an outcome.

Use your source notes to check facts and preserve qualifications. An opening should lead into the material you actually show. You can write alternatives yourself or request generated candidates from reviewed notes. Neither path ranks hooks or predicts viewer performance.

Compare, review and apply.

1. Open a video and expand Compare opening variants.
2. Write two alternatives, each with up to 240 characters of main wording and an optional supporting line up to 160 characters. Add a third when it offers a distinct approach.
3. Select Compare written openings. Check the current first scene against the alternative wording and choose an opening.
4. Review its facts, qualifications, reading time, existing narration and timed captions. Confirm that review before selecting Use selected opening.
5. Preview the first scene, save a version, export, and watch the completed video. One Undo restores the previous wording.

The first scene retains its duration, appearance, background and graphics. Other scenes, music, narration, source links and timed captions stay as they were. New wording can disagree with an older recording or subtitle track: correct those separately before exporting. Use the script timer or teleprompter to check reading time.

Generate candidates from reviewed notes.

Save the current video, then expand Generate opening candidates inside the opening comparison. Choose English or Spanish, describe the opening goal, add source notes with their qualifications, and confirm permission to use those notes.

The model receives your notes, title and current opening wording. It does not inspect the media, look up facts or verify claims. It drafts two or three short alternatives within an approximate first-scene word budget. The video stays unchanged.

Completed candidates and their source notes remain in workspace history. Select Compare these generated openings, then choose and review one before applying. Generated candidates must match the exact saved version and complete current draft. If either changes, earlier candidates remain readable but cannot be loaded into that comparison.

The trial allows two opening attempts per workspace per UTC day, including failed or cancelled attempts, and one active request. Shared capacity may also limit admission. Keep an uncertain request’s original body and key: Retry same opening request recovers its confirmation without deliberately starting another inference. Check retained attempts before clearing a request. An uncertain model result is not retried automatically.

A queued attempt can be cancelled. A model call already running may have incurred cost, but late results are discarded. Retained review and cancellation remain accessible when new generation is disabled.

Keep alternatives and separate experiments.

Written choices stay in the current visit and are cleared when you apply one or discard them. Download written openings keeps a plain-text copy of the current opening and alternatives before you apply. It does not include private media. Leaving with unfinished choices prompts you to keep editing or discard them.

Comparison pauses autosave. Applying leaves autosave off so you can inspect the wording and soundtrack before saving. If the video changes after comparison, compare again; the app keeps your alternatives and blocks the earlier review.

If word-removal edits hide the entire first scene, restore that removed time before comparing its opening. Changing wording on an invisible scene would not change what viewers see.

For independent versions, feedback and exports, first save the current video and choose Duplicate video. The copy retains a link to its saved origin and reuses private workspace media. Apply an opening to the copy, then follow the version review workflow. A copy does not inherit an approval or client preview link.

A wording comparison is a starting point for an experiment. It is not evidence that one hook performs better. Any later comparison needs actual publication results and a clear account of what else changed.