The first thing to know about export settings is that nobody ever watches your export. Every platform re-encodes what you upload — always, whatever you send — so your file is not the video people see, it is the master their compressor works from. That reframes the question. There is no magic number that makes TikTok favour you; there is only a range: enough resolution and bitrate that the platform’s encoder has clean material to work with, and not so much that you are uploading gigabytes it will immediately throw away. Both mistakes cost you. Undershoot and the platform’s compression amplifies your compression, stacking artefact on artefact. Overshoot and nothing is gained except upload time — a 4K file sent to a feed that caps playback at 1080p is downscaled before anyone sees it.
The caps differ by platform, and by kind of platform. TikTok and Instagram Reels are vertical feeds: 9:16, and 1080×1920 is the practical ceiling — they will take bigger files, but what plays back is 1080p-class, so that is the master to send. YouTube is the opposite: it keeps 4K, and higher-resolution uploads tend to be given the platform’s better compression, so if the take was shot in 4K it is the one place that resolution pays. LinkedIn sits in between — a feed, mostly watched at 1080p on a phone, forgiving about aspect ratio. Bitrate follows the same logic: for 1080p, something in the region of 8–12 Mb/s in H.264 is already more than any of these platforms preserve. Duration is the quieter constraint — each feed has its own tolerance, and it is worth respecting before you export rather than after.
Keeping four sets of caps in your head is exactly the kind of job software should do. Talk2Camera ships a preset for each of the four platforms, each carrying its own resolution cap, bitrate and duration guidance — the vertical feeds get a 9:16 frame, the others keep the shape you shot. The export is also honest about work it does not need to do: a plain trim that already matches the preset copies its frames instead of compressing them again, which is faster and loses nothing. And when it finishes, it tells you what actually happened — whether the video was re-encoded or copied, at what size, and whether the frame was cropped — and a long export can be cancelled mid-run rather than waited out. You should never have to guess what a preset did to your pixels.
Generation loss is why the copy-instead-of-re-encode detail matters more than it sounds. A video codec compresses by discarding what it judges you will not notice, and the judgement is made fresh on every encode. One generation is genuinely hard to see. But each further encode judges the previous encode’s output, so the discards compound: skin flattens, gradients band, shadows go blocky. Your upload will always be one generation older by the time it plays — the platform’s encode is unavoidable — so the goal is to reach the platform in as few generations as possible, ideally one: theirs. Practically: export from the original project, not from a previous export; keep the highest-quality master you have; and when a tool offers to copy frames untouched because nothing needs changing, take it every time.