Timelapse Pipeline: Frame Extraction, Assembly and Speed Math
Two hours of drifting clouds compressed to thirty seconds, a five-minute construction job fast-forwarded to one — timelapse needs no special gear, just one video and two processing steps.
The principle: extraction interval is time compression
Timelapse keeps only the key moments: a 30fps source holds far more frames than the effect needs. The extraction interval is the granularity of time compression — one frame every 5 seconds turns a two-hour video into a 1,440-frame pool, which is exactly the raw material of the effect.
Two variables set the speed ratio: interval and per-frame duration. Output duration = source duration / interval x per-frame duration. Example: a 10-minute (600s) source at one frame per 5 seconds yields 120 frames; at 0.5s each, the output runs 60 seconds — a 10x timelapse. Raise the interval or shorten the dwell to go faster.
Extraction: rate and format
Match the rate to content speed: slow motion (clouds, light, construction) is fine at one frame per 5 seconds; street traffic and cooking want one per 2 seconds; motion analysis needs multiple per second. Too high a rate just fast-forwards the original — the timelapse feel comes from sparse sampling.
Format follows the downstream: JPG (quality factor 2) for direct assembly; lossless PNG when frames will be retouched first. A long extraction produces thousands of images — mind your disk space.
Assembly: dwell time and frame width
Assembly dwell starts at 0.5 seconds — the lowest tier is the fastest rhythm. For sub-0.2s dwell you want video speed change, not frame extraction: two routes, smooth acceleration via a speed tool versus the signature stepped rhythm of timelapse assembly.
Output width defaults to 1280 with height following the first frame ratio. Publish-ready files go through one more compression pass. Check two things in the result: first and last frames present (interval boundary behavior) and even rhythm (consistent intervals guarantee it).