A documented 15-second 1080p render on Seedance 2.0 took roughly 20 minutes during heavy platform load — treat that as an upper bound, since queue load makes render times variable. Dropping to 720p cuts generation time dramatically, and image generation now completes in seconds. End-to-end, a finished AI ad typically takes 2–3 hours because of iteration.
Budget roughly 20 minutes for a long 1080p clip and considerably less for short ones: the one documented 1080p benchmark is a full 15-second multi-shot video rendered in a single Seedance 2.0 pass in about 20 minutes — and that was during a period of platform instability, so it sits at the slow end of the range. Render time is not fixed; the same clip can take different amounts of time depending on how loaded the generation queue is, so plan generation runs with slack rather than assuming a constant per-clip number.
Resolution is the biggest lever you control. 720p renders significantly faster than 1080p on Seedance 2.0, and 720p is visually sufficient for most vertical short-form delivery — so iterate at 720p and reserve 1080p renders for clips your delivery format actually needs at full resolution. invideo is an agentic video creation tool with all the current video models — Seedance 2.0, Kling, Veo — available in one place, so the invideo agent can route each clip to the model that fits the shot without you switching platforms.
Still-image generation is far faster than video: after recent platform updates it completes in seconds where it used to take minutes. That speed difference is why the fastest workflows iterate on a still frame until the framing is locked, and only then spend a video render animating it — you burn seconds per iteration instead of many minutes.
Per-clip render time is also a fraction of real production time, because most generated clips get discarded. In one documented production, 39 video clips were generated to yield 10 usable ones (26% utilization); in another, 1 of 11 generated clips made the final cut. Generating in batches of about 5 clips keeps that iteration loop fast and catches problems before a large run goes wrong. Accounting for iteration, documented end-to-end times run ~2 hours per finished UGC ad (after a one-time 15–20 minute brand context setup), ~2–3 hours per product ad, and 3–4 hours for a full campaign of 40 stills and 30 motion clips — so 2–4 hours per finished deliverable is the realistic planning number, even though any single 1080p clip renders in minutes.
Watch some of these to see what works for you:
We're talking seconds for things that used to take a couple of minutes. So, that's awesome from a workflow perspective.
— invideo's creative team, on image generation speed after recent platform updates