YouTube Shorts length and resolution limits
Shorts cap at three minutes and 1080p. A hard delivery ceiling changes what generation resolution is for, and turns downscaling into a quality decision.
Two numbers define the Shorts container and one of them is a ceiling rather than a target. YouTube's own documentation is short about it: Shorts are short-form videos up to three minutes long, and you can upload Short videos with a maximum resolution of 1080p.
The three minutes get all the attention. The 1080p is the more interesting constraint, because a hard ceiling changes what generation resolution is for. When delivery is capped, rendering above the cap stops being a delivery decision and becomes a supersampling decision, which is a completely different trade with a completely different answer.
The spec, first
| Property | Value | Notes |
|---|---|---|
| Maximum length | 3 minutes | Claimed content is blocked over 1 minute |
| Maximum upload resolution | 1080p | Documented ceiling, not a recommendation |
| Orientation | Square or vertical | Upload docs accept both; compose 9:16 anyway |
| Working canvas | 1080 × 1920 | 9:16 at the resolution ceiling |
| Frame rate | 25 fps by default in our pipeline | Nothing in the Shorts spec requires a specific rate |
Two of those rows carry consequences beyond the number.
The three-minute ceiling interacts with claims. Shorts over one minute in length that contain claimed content are blocked. Under a minute, a claimed track can still run; over a minute, it cannot. Royalty-free tracks from YouTube's Audio Library are documented as not receiving a Content ID claim. If you are planning a two-minute Short, plan unclaimed audio at the same time, not afterwards. This is a common way a longer Short fails after it is already assembled.
1080p is a maximum, and the documentation states it as one. That is different from a recommendation. Treat anything above 1080p as something you resample yourself before upload rather than something you hand to YouTube and hope about.
What a delivery ceiling does to your generation settings
Here is the decision that actually matters, and it only exists because the ceiling exists.
Generate at 1080p and you deliver what you rendered, one pixel to one pixel. Generate above 1080p and downscale, and every delivered pixel is an average of several rendered ones. That averaging is not neutral. It suppresses per-pixel noise, softens the high-frequency shimmer that generative video models produce in fine texture, and cleans up edge stepping. The output is typically calmer than a native 1080p render from the same model and prompt.
So on a platform with a 1080p ceiling, higher generation resolution buys you quality at 1080p rather than a higher-resolution deliverable. It is a denoising pass you paid for in generation credits.
Whether that is worth it depends on three things:
- Does the model have a native vertical mode at or above 1080p? If yes, that is usually the cleanest path. Native vertical avoids a reframe step entirely, and the argument for generating vertical natively rather than reframing afterwards applies regardless of resolution.
- What is the model's best-quality mode? Some models are demonstrably better at one internal resolution than at others, and it is not always the highest one. If the sharpest, most temporally stable output comes from a mode above 1080p, generating there and downscaling converts that headroom into visible quality at the ceiling. If the model is at its best at 1080p, going higher spends credits for nothing.
- How much texture is in the shot? Downscaling helps most where shimmer lives: hair, foliage, fabric weave, crowd detail, water, fine typography. On a flat-graphic explainer or a simple talking head against a plain background there is little for the averaging to fix.
The honest summary is that the supersampling gain is real and it is not large, and it is not worth paying for on every Short. It is worth paying for on the ones where the model is visibly struggling with texture, which you can identify by looking at one render rather than by rule.
The case below the ceiling
The opposite situation is more common than people expect: a model whose best or only vertical mode lands below 1080p. Now you are choosing between delivering under the ceiling and upscaling to it.
Upscaling to reach a ceiling is the weaker move, and the reason is specific. An upscaler does not recover detail that was never rendered, it synthesises plausible detail. On stills that is often fine. On video it introduces a second temporal-consistency problem on top of the model's own, because the synthesised detail is regenerated per frame and does not always agree with itself between frames. What you get is a sharper-looking Short that crawls.
Delivering a clean 720p render inside a 1080 × 1920 container is frequently the better outcome than a hallucinated 1080p one, especially on a feed that is watched on phones at speed. Whether to upscale the render or generate at full resolution works through that trade properly, and two-pass regeneration versus native high resolution covers the third option, which is to re-run the generation rather than to enlarge it.
The export spec
Assuming you have made the resolution decision, the rest is mechanical.
- Canvas: 1080 × 1920.
- Orientation: vertical, composed for vertical. A landscape shot letterboxed into a vertical frame can meet the aspect-ratio rule and still die in the feed.
- Frame rate: consistent across the timeline. Mixing sources is where problems start, and 24 fps clips on a 25 fps timeline is the specific failure mode to know about if your generated clips and your stock footage disagree.
- Resample once. If you are downscaling from a higher render, do it in one step on the finished timeline rather than per clip. Repeated resampling compounds softness.
- Safe zones: the Shorts player draws its own chrome on the frame regardless of resolution. Keep burned-in text off the right-hand icon rail and the bottom band that holds captions and audio credits.
- Audio: decide whether the track will be claimed before you lock length, per the one-minute rule above.
If you are converting existing landscape material rather than generating vertical, changing the aspect ratio in the editor is the step where composition gets decided, and it deserves more attention than the resolution question does. A correctly-resolved badly-framed Short loses to a softer well-framed one every time.
A decision rule
Three questions, in order:
- Is your delivery canvas 1080 × 1920 vertical? If not, fix that first. Nothing below is worth doing on a wrong canvas.
- Does your model produce a clean 1080p vertical render? If yes, render there, deliver there, stop. Do not spend credits reaching for a resolution the platform will not accept.
- Is the model shimmering on texture? If yes, and a higher mode exists, render higher and downscale once at the end. If no higher mode exists, prefer the clean lower-resolution render over an upscale.
That is the whole thing. The ceiling makes it a much shorter decision than it is on platforms without one, which is the underrated benefit of a documented maximum: it removes an entire axis of speculation. There is no resolution arms race available on Shorts. There is only how good your 1080p is.
Resolution as a concept and our resolution guide cover the general case if you are choosing generation settings across several platforms at once, where the ceilings differ and the answer stops being this simple.
FAQ
Should I generate at 4K and downscale for Shorts?
Only if your model's output at that resolution is genuinely better and the shot has texture that benefits from the averaging. The delivered file is 1080p either way, so the higher render buys quality rather than resolution, and it costs more credits. On flat graphics and simple compositions the gain is close to invisible. On hair, foliage and fine fabric it is worth looking at.
Does a three-minute Short still behave like a Short?
The length limit is a container rule and running to three minutes is allowed. The audio constraint is the practical one: Shorts over a minute containing claimed content are blocked, so longer Shorts need audio that will not receive a claim — original score, Audio Library tracks, or other unclaimed audio. Beyond that, the format question is whether your idea holds three minutes, which is an editorial decision and not a spec one.
What frame rate should I export at?
The Shorts spec does not require a particular rate. What matters more is internal consistency: one rate across the timeline, chosen before you start mixing generated clips with anything else. Our default is 25 fps, and the failures we see come from mixed-rate timelines rather than from the rate itself.
Is square accepted, or does it have to be 9:16?
YouTube's upload page accepts Shorts with a square or vertical aspect ratio. Compose for 1080 × 1920 regardless. The feed is a full-height vertical surface and anything short of that is giving away frame for no reason.