A vertical clip is not stored as one file. Feed-based playback requires several renditions of the same short for reasons that have little to do with connection speed.
Feeds preload the next clip
Scrolling feeds must start the next video instantly, which means beginning its download before the viewer has decided to watch it.
Preloading every upcoming clip at full quality would consume enormous bandwidth for material most viewers skip past within a second.
Apps therefore fetch a small, low-quality opening portion of upcoming clips, and switch to a fuller rendition once a viewer stays.
The first moment is encoded separately
Because the opening fragment is what most viewers see, it receives disproportionate attention in the encoding pipeline.
Some pipelines encode the first seconds at higher quality relative to the rest, on the reasoning that the decision to keep watching is made there.
This produces the effect viewers occasionally notice, where a short looks its best at the start and settles into something slightly softer afterwards.
Aspect ratios multiply the renditions
The same short may be shown full screen on a phone, in a smaller frame on a tablet and letterboxed on a desktop browser.
Rather than scale one file for all of them, platforms often produce crops and renditions matched to the surfaces where the clip will actually appear.
A creator who frames content tightly can lose important elements in one of those crops, which is why safe areas matter more in vertical than in horizontal work.
Loops interact with encoding
Short-form clips replay automatically, so the transition from the final frame back to the first is seen repeatedly.
Players handle this by keeping the decoded clip available rather than re-requesting it, which is why a loop is seamless while the first play may not be.
Encoding a clip so its ending and beginning match visually removes the jolt at the loop point, an editing decision rather than a technical one.
What creators should upload
Since every rendition is derived from the upload, the only lever a creator holds is the quality and framing of that source.
Uploading a heavily compressed export gives the pipeline degraded input, and no downstream rendition recovers what was already discarded.
Uploading at the platform's stated maximum, in its preferred aspect ratio, costs only upload time and improves every derived version at once.
Exporting with a generous bitrate matters more than exporting at a higher resolution, because it is the bitrate that determines how much detail survives the first encode.