Based on the image, I'm filling the 7 drone variables:
- SUBJECT: A young woman sitting on a black bollard
- OUTFIT: White tank top, black oversized cardigan, ripped light-wash jeans, black sneakers
- ACTION: Sitting casually, one leg up on the bollard, relaxed pose
- BACKGROUND: European cobblestone street at dusk
- FILM_STYLE: Fujifilm Pro 400H (matching the moody, slightly desaturated tone of the original)
Since this is a single source image to animate, I'll use image-to-video mode — the frame locks the appearance, so I only describe the motion and camera move.
Cinematic drone reveal: starting from the portrait-level frame, the camera will crane upward and arc around the subject, settling on a high-angle front view — giving the street portrait a sweeping, cinematic feel with the moody Fujifilm color grade preserved from the original.
References
Video is rendering now — Street Portrait Drone Reveal (16:9, 10s, 720p). It'll finish on its own in a minute or two.
The shot will crane up and arc around your subject on the bollard, settling into a high-angle front view, with the moody Fujifilm Pro 400H color grade matching the original street portrait. The result will appear on the canvas automatically — no need to refresh.
I see the new image — a different street portrait with a subject in a black hoodie and blue cap on the same bollard setting. What would you like to do with this one?
You are right. The video generation did not fully replicate the girl's appearance in the image. I will regenerate it, this time placing greater emphasis on maintaining consistency in the facial features.
Regenerate the drone reveal, again with frame #1 as the starting point, but this time emphasize in the prompt that the character's face, facial features, and overall appearance must remain completely consistent without any changes.
References
Okay, regenerate in vertical 9:16 format.
Regenerate drone reveal, vertical format 9:16, emphasizing maintaining exact facial consistency of the person, with #1 as the first frame.
References
The vertical version is currently rendering — Street Portrait Drone Reveal v2 - 9:16, 10 seconds, 9:16 vertical, Fujifilm Pro 400H color tones.
This time, the prompt specifically emphasizes maintaining complete consistency in the subject's face and overall appearance. Once generated, it will automatically appear on the canvas, and you can check the fidelity of the restoration.
The vertical version of the drone reveal video has been generated. You can play and view it on the canvas. This time, I emphasized keeping the facial features consistent in the prompt. Let me know if you're satisfied with the resemblance or if there are any adjustments needed. Feel free to tell me anytime.