A 4:3 image gives the Goodbye GTA 5 AI video room for four people on a pier. A vertical crop can cut the people at either edge out of the scene. The same crop can look perfectly fine a few seconds later, when the shot cuts to a close-up of the driver. That inconsistency is the reason to treat framing as part of the export design.
Keep the full 4:3 frame inside the 9:16 canvas
At 720 pixels wide, a 4:3 frame is 540 pixels tall. A 9:16 canvas with the same width is 1,280 pixels tall. Fitting the landscape frame inside that canvas leaves 370 pixels above and below it. Every person in the original shot stays visible, though the main image occupies only about 42% of the vertical screen.
A solid background makes a good baseline for testing. The variant in the case study fills the extra space with a cropped, blurred copy of the same frame. The blur belongs on the background layer. The centered foreground carries the original framing. On a phone, keep an eye on how much visual detail the background adds during bright cuts or fast movement.
Start with this FFmpeg export
This example uses a 720×1280 canvas and a dark background. It maps the first audio track when the source has one, and encodes the output as H.264 video with AAC audio.
ffmpeg -i source-4x3.mp4 \
-vf "scale=720:1280:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=720:1280:(ow-iw)/2:(oh-ih)/2:color=0x110d0b,setsar=1" \
-map 0:v:0 -map '0:a:0?' \
-c:v libx264 -pix_fmt yuv420p -c:a aac \
-movflags +faststart vertical-9x16.mp4
For a blurred background, make a second branch from the source, scale it to fill the vertical canvas, crop it, then blur it before overlaying the fitted foreground. Both branches use the same source frames, so the background follows every cut in the foreground. The original audio can take the same path as in the example above.
Let the two downloads become ready independently
The 4:3 render is the source for the vertical copy. That makes the 9:16 export a separate job. An order page can show the original download as soon as it is ready and show a pending state for the vertical copy. If the second job needs a retry, the existing 4:3 file can stay available. A stable output key for each format makes it easier to retry without creating duplicate downloads.
Before calling the vertical export done, check the dimensions, the display aspect ratio, and whether the expected audio track survived. Then watch the group shot and the close-up on a phone. A stream probe catches the dimensions and the track list, but viewing the file exposes framing or sync problems a metadata check cannot reveal.
The case study is GTA 5 Goodbye AI, my one-photo video project. The public sample includes the pier and car shots discussed above, which makes the framing trade-off easy to see. The sample is on gta5goodbyeai.com.
Top comments (1)
Some comments may only be visible to logged-in visitors. Sign in to view all comments.