Unstable detection of transparent areas

[mimimi999]

I'm currently testing various prompts, and I'm seeing the following results most often:

・Only a portion of the background is transparent.
・The foreground is also partially transparent (e.g., human black hair or shadow areas).

I believe this is due to a discrepancy between the content of the training videos and the generated videos. Are there any prompt words that, if included during generation, would increase the success rate of background transparency?

[daooshee]

Thank you for your interest in our work! Can you share the prompts that have unsatisfactory results?
Our model has specific prompt requirements. For example, you need to specify that the background of the video is transparent, the visual style, the shot type (such as close-up, medium shot, wide shot, or extreme close-up), and a description of the main subject. An example is "This video has a transparent background. Close-up shot. A colorful parrot flying. Realistic style."

[mimimi999]

Sorry, I overlooked the prompt requirements in the guidance explanation. By entering the prompts as instructed, the transparent video was generated without any problems.