Maya has three polished AI tool demos open and no answer to a practical question: which workflow suits her own candle? In this fictional holiday project, she runs a handmade candle shop. She shoots on iPhone and edits in iMovie on her Mac. None of the demos shows her amber glass candle or its handwritten label.
Her AI video generator test follows the candle from the source photograph to the final phone preview. She will fix a brief, prepare an iPhone test kit, check each candidate’s purpose, compare compatible outputs, then verify the Mac handoff and publishing permissions.
Write a Brief for One Candle Clip
Save one sentence in Notes: “Make a 15-second, 9:16 candle video for Reels, with the handwritten label readable.” This is Maya’s creative target, not a claim about any model’s output options.
Add a short acceptance note. The amber glass must still resemble the actual product. The wooden lid must move in a way her real candle permits. The label’s words must remain correct in the portion she intends to use.
Keep the same brief for each candidate. Reject a clip that fails a required product detail.
Gather the Same iPhone Test Assets
Prepare an authorized product photograph, a short iPhone clip, and a three-line script. For Maya’s sample kit, the recording runs between ten and twenty seconds and shows the wooden lid being lifted. The photograph gives her a clear reference for the handwritten label and glass color.
Put the photo and clip in a dedicated Photos album. Keep the script in Notes, and add a screenshot of those lines to the album if that helps keep the kit together. Preserve the originals separately from generated outputs.
A proposed script might identify the candle, introduce the lid-opening action, and direct attention to the label. Keep that wording constant. For the image-to-video test, submit the approved photograph and the same movement instruction. Keep the iPhone recording offline as the reference for the real lid action; it is not a required model input.
Illustrative test kit for Maya’s fictional candle: the still preserves the label and glass color, the original phone clip shows lid motion, and the blank script stands for wording to keep constant. No generator result is shown.
Check What Each AI Video Generator Makes
Read the candidate’s own description of its inputs and output.Wizstar AI is one vendor blog with model explanations to consult alongside the candidate’s documentation. Its Jev article describes a model for structured decisions that does not generate video or images. That is a category distinction, not a quality judgment or a video test result.
Apply the same distinction to the job at hand. A presenter-focused workflow, an enhancement tool for existing footage, and a generator for new product visuals answer different requests. Check which part of Maya’s project the candidate actually supports.
Record the supported route. Compare image-to-video candidates using the same photograph and scene instruction, keeping the script for assembly if direct script input is unavailable. A photo-only route stays eligible for this trial; it does not need to ingest the reference recording. Test any video-input workflow separately so its results are not treated as the same experiment.
Compare Clips Against the Same Brief
For compatible image-input candidates, submit the same approved photograph and movement instruction. Use the same script when assembling each draft. Start with the defaults and record required setup. Give each candidate the same revision allowance instead of polishing one repeatedly and judging another by its first output.
Download the drafts and view them on the Mac with the source files nearby. Compare the same action or selected segment rather than relying on the promotional preview. Use the real iPhone product video as the reference for the lid movement.
| Check | What Maya inspects | Record after the trial |
| Label | Exact wording and letter shapes | Defect and where it appears |
| Lid action | Hand contact and opening movement | Difference from the real clip |
| Product | Amber glass, shape, wooden lid | Any changed detail |
| Brief | Required story and usable segment | Meets requirement or needs work |
Leave the result fields blank until the trial. Watch the retained sequence, including intermediate frames. A correct label at the start needs to stay correct later. Check whether a hand passes through the lid or the glass changes its apparent shape.
Write one concrete observation per check. “Label word changes halfway through” gives Maya something to investigate. “Looks better” leaves the decision unclear. Keep rejected drafts. Note any extra work a version needs. The record supports a decision about this candle task, without establishing general tool superiority.
Test the File in Your Mac Workflow
Inspect the downloaded file before committing to an AI video tool for Mac work. Check its actual format, dimensions, and any visible watermark. Trial and paid-account conditions need checking separately; assume nothing about what an upgrade removes.
Try importing the file into the current iMovie project. Apple documents importing iPhone photos and videos, but that says nothing about whether every generated file will work with the same workflow. Play the imported clip and check it against the downloaded version.
Add the titles and appropriately licensed music the project needs, then export a trial cut. This iMovie workflow is part of the test. A file that plays in a browser still needs to survive the editing and export steps.
Check the exported file’s width, height, and duration in its file information. Width divided by height must equal 9 divided by 16 for Maya’s target; a vertical preview alone is insufficient. Mac iMovie’s documented cropping behavior includes a 16:9 frame. If this edit exports the wrong shape, open a verified editor that offers a 9:16 canvas, import the reviewed clips, reframe the candle, and reposition titles before exporting again. Recheck the new file’s dimensions and 15-second target.
Review the final file on iPhone and in the intended posting preview. Check the label, caption edges, crop, and selected length again. A later crop can hide the detail that passed the earlier review. Record the workable handoff alongside the visual checks.
Save the downloaded draft and final export separately, with filenames that identify the candidate and version. If a file fails at import or export, note the stage and retain it for checking instead of quietly replacing it with a different test.
Verify Rights and Save the Test Kit
Read the candidate’s current terms for the intended commercial use. Check the source material and music permissions too. An accessible export is not evidence that every element is cleared for a shop’s promotion.
Review the publishing platform’s current requirements for AI-generated or modified content. Describe the contribution accurately wherever the applicable process requires it. Keep the shop test focused on its own product footage.
Maya’s next step in this hypothetical workflow is to choose from the recorded evidence, if a candidate meets her requirements. She keeps the original kit, brief, script, and results for the next model change.
Build that kit today: one authorized photo, one short clip, and one three-line script, with the media collected in a dedicated Photos album.
