The reason the estimated video duration keeps changing is because of an applied voiceover, and you're likely to notice a difference in the estimated video duration after you preview your video.
When you apply an AI voiceover, it doesn't become part of the video until you preview it. During the preview, the video duration may change based on how long it takes the voiceover to read the text in each scene.
In a case where you are using a recorded or uploaded voiceover in the entire video, then the video duration will be synced to the length of the audio as well.
After previewing, the 'estimated video duration' will be updated to reflect this change. However, the duration you see in the preview will be the same as the duration in the downloaded video, unless you change the voiceover again.
In essence, the actual video duration is only set after you apply a voiceover and preview the video.