This is a republication of Reedo’s Naver article. It is not a new experiment or a DMS client case study. The cover and in-article photographs are AI-generated staged scenes, not evidence of actual work.
When you create a product video with AI and the bottle changes shape in the next result, it can be hard to know which words to change. In this Google Flow Short, the creator generated three versions of each of three prompts. A short prompt that explicitly specified a cylindrical shape produced a cylindrical bottle all three times.
That does not mean a shorter prompt alone will produce the same product. The prompts differed not only in length but also in the words they included. It is more accurate to treat these results as a starting point for checking whether you left out the features you want to preserve, rather than as a formula for the perfect prompt.
The key is to specify the shape you want to preserve and check the repeated results separately, rather than simply writing less.
AI-generated staged image, not an actual work screen.
What was actually compared?
The first prompt described a blue glass perfume bottle, a light gray table, and slow camera movement, but did not specify the bottle's shape. The second gave a longer description of lighting, reflections, shadows, and even the camera angle, but again included no word defining the shape. The video reports that under both conditions, the bottle shape differed in all three generations.
The third prompt included the word cylindrical, a navy blue cap, a white table, and forward camera movement. Although the proportions varied slightly, all three results showed a cylindrical bottle with a navy blue cap. The long prompt also preserved the navy blue cap and white background, so it would be wrong to conclude that all of its instructions failed.
The shared settings listed in the description were Omni 1.1 Flash, vertical 9:16, 360p, four seconds, and one output per request. There was no starting image, and the creator stated that they could not find a seed-locking feature. Keep the scope in mind: three generations for each of three prompts, for a total of nine.
AI-generated staged scene 2View original
AI-generated staged image, not an actual work screen.
Why can't we say shorter is better?
The effect of adding the word cylindrical was not separated from the effect of shortening the prompt. The table color and wording for the camera also changed. These results alone therefore cannot establish that prompt length determines consistency, or that specifying a cylindrical shape makes every product come out identical.
The video also reports that the camera moved much closer with the short forward-movement instruction, while it barely moved with the longer prompt that included an angle. This, too, is an observation under those particular settings. With only one object and one model tested three times per prompt, it is best not to assume that other materials or movements would show the same differences.
AI-generated staged scene 3View original
AI-generated staged image, not an actual work screen.
What should you check in your own practice?
The following are suggested checks for your own comparisons, not results from additional experiments in the video. First, describe the bottle body's shape, cap color, background, and movement separately. Defining visible features such as a cylindrical body and a navy blue cap, rather than subjective descriptions such as a beautiful product, gives you clearer criteria for comparison.
To examine the effect of length, keep the key shape words unchanged and vary only the additional description. It is also worth recording the model, aspect ratio, duration, and whether a starting image was used. Rather than picking only one good result, retain every result from your planned number of attempts so you do not overlook conditions that failed.
When reviewing the results, distinguish between having the same cylindrical shape and being exactly the same bottle. Check separately whether the body proportions and cap size match, and whether the silhouette changes during the video. Sharing the cylindrical category does not guarantee an identical product design, so set your acceptance criteria to match the needs of the scene you intend to use.
Watch the actual process in this video
This Short does not show that detailed descriptions are useless. It presents a limited comparison of what happened when the desired shape was stated explicitly. Start by identifying missing features, then change one condition at a time and review all the results. Defining precisely what you mean by the same bottle will make it clearer why you should revise your next prompt.
