ad
ad
Topview AI logo

27 - AI Background Remover, Does It Work?

People & Blogs


27 - AI Background Remover, Does It Work?

In this article, we'll delve into another AI feature of CyberLink's PowerDirector 365—the AI Background Remover. While previous videos have been somewhat critical of CyberLink's AI features, we aim to dissect this one and see how well it performs.

Using the AI Background Remover

We start with the launcher program and click on the AI Background Remover applet. Although you can drag any video into the input box for the AI, this can also be easily accomplished within the program. Instead, we'll click on the option to try with a sample clip. The main program now loads with the sample video in track two and a color board on track one to supply some color to the transparent background.

Zooming in on the timeline makes it easier to see what's happening, and expanding the height of the track helps to see it more clearly. The clip is just eight seconds long.

To understand what the AI does, we reverse engineer the clip. By deleting the color board on track one, we can see the transparency. Upon closer examination, hovering the cursor over the small green box at the beginning shows "Tools" and "AI Background Remover." Clicking on the green box and then on the AI Background Remover shows us that the transparency option was selected. Removing the check mark returns the clip to its original state, and playing it reveals all objects in the background.

Critiques and Advanced Options

From this examination, we can say that the AI Background Remover did a good job with the sample clip. However, we noticed that the menu for AI only has two choices: a transparent or a blurred background. There is also an "Advanced" button that seemingly suggests more options, but clicking it reveals there are no additional features. This raises the question of why there's an "Advanced" label if nothing more is provided.

While we are in this window, switching to a blurred background and selecting the maximum setting shows that the effect is applied instantly and well. Returning to the timeline and hitting play confirms that the AI performs excellently in these two scenarios.

Testing with Real-World Videos

To really test the AI background remover, we use our real-world videos. First, we take a clip named “with hat.” The subject is a gentleman in costume for a New Year's Eve appearance. After applying the AI, we notice glitches around the hat, making the feature far from intelligent or acceptable.

Switching to a blurred background shows improvement but glitches remain. The AI, it seems, isn't aware that humans can wear hats.

Next, we try a clip named "no hat." The subject has multicolored hair—a more likely candidate for success. Applying the AI Background Remover to this clip results in acceptable performance with only minor glitches around the gray hair. Blurring the background further smooths out these issues to a degree of acceptability.

Conclusion

While the AI Background Remover works excellently in controlled and high-contrast environments, real-world scenarios expose its limitations. That said, it shows promise, and with some refinement and code updates, it could become a very practical tool.

Keywords

  • AI Background Remover
  • PowerDirector 365
  • CyberLink
  • Transparent Background
  • Blurred Background
  • Video Editing

FAQ

Q: What are the options available in the AI Background Remover? A: You can select either a transparent background or a blurred background.

Q: How does the AI Background Remover perform with a sample clip? A: It performs well, handling stationary subjects in high-contrast environments effectively.

Q: Are there any advanced options available for AI Background Remover? A: No, the "Advanced" button does not reveal any additional features.

Q: How does the AI feature handle real-world videos? A: The AI struggles with more complex, real-world scenarios, especially involving objects like hats or intricate hair patterns.

Q: Is the AI Background Remover usable for professional editing? A: It still needs refinement but shows promise in specific, simpler scenarios.