You will fail if you don’t read the following manual, Please use an image of 1 standing woman, showing above the waist.
Please use an image with the following composition:

・Only one person in the image
・Above the waist only is visible
・The entire area above the waist is fully included in the image
・The person is facing forward and standing

Images of sitting poses will lower the success rate, while images of lying poses will almost certainly fail.
The highest priority is to ensure that the face does not turn into a different person.
Therefore, please understand that the success rate will drop significantly when using any composition other than the specified one.

Cropping the image to match the specified composition will increase the success rate. (it’s easy to do by taking a screenshot on your smartphone).
However, please note that the following types of images will lower the success rate:

・Images created or heavily edited by AI
・Images that include unrelated people
・Images with angled poses, leaning, lying down, sitting, or not facing forward
・Images that are too bright or too dark
・Images with reflections or mirror glare on glass or mirrors
・Images depicting minimal-coverage outfits that AI may have difficulty identifying as regular clothing.

Such images are difficult for the AI to analyze correctly, which increases processing time and lowers the success rate.
In short, the more visual information an image contains, the lower the success rate will be.
There is absolutely no need to use high-resolution images — they only reduce the success rate.

If you want to create freely on your own, try using “Try400–800”.
To create a longer video, take a screenshot of the last frame and crop it. Then, use “Try400–800” to generate the next segment. Repeat this process, and finally, connect all videos using free software like “Microsoft Clipchamp.”

The cost is automatically calculated by SeaArt based on factors such as generation time, image size, and step count.
It is not determined by the creator of the app (myself).
If you plan to use it repeatedly, running it during off-peak hours may help slightly reduce the cost.

Use and share the generated results at your own risk.
The creator is not responsible for any issues or disputes arising from the generated content.

コンテキスト拡張

評価とレビュー