Clip a video¶

Create a new video from a section of an existing, completed video: a trim or highlight.

Both endpoints cut the same way and give the same picture (see How clips are produced). They differ in what the clip stores:

• clip: an independent video. It copies the frames it needs into its own media, so it stands on its own and the source can be deleted afterwards.
• fastclip: much faster and far smaller. Rather than copying, it keeps referencing the source's already-encoded segments — so the source cannot be deleted while fast clips of it exist.

Choosing the range¶

Specify the cut with start_time and end_time, in seconds:

{ "metadata": { "title": "My highlight" }, "start_time": 10, "end_time": 50 }

metadata optionally carries the clip's title and public flag. Encode-time settings (watermark, encoding tier, audio normalization) are inherited from the source video: the clip's media derives from the source's already-encoded renditions, so they can't be chosen per clip.

How clips are produced¶

Your video is stored as short (~6s) independently-decodable segments. For your cut, both clip and fastclip:

• keep every whole segment the cut spans, exactly as encoded, not re-encoded; and
• re-encode only the two boundary segments (the one the cut starts inside and the one it ends inside), and even then only the partial frames around the cut point.

So the amount of re-encoding is small and roughly constant whatever the clip's length (a 5-minute clip re-encodes about as much as a 30-second one), the middle of the clip keeps exactly the quality you already encoded, and the cut lands on the frame nearest the times you asked for — on the frame grid all quality levels share, so every level starts the clip on the same instant (for a 60fps video whose lower levels play at 30fps, that grid is the 30fps one).

The two endpoints part ways only at the last step. fastclip stores just the re-encoded boundaries and points at the source's segments for everything between them — which is why it is so much smaller and faster, and why it ties the clip's life to the source's. clip copies those segments into media of its own, giving a fully independent video.

If a cut is too short or misaligned to enclose a single whole segment (say, under ~6s), there is nothing to reuse and the video is re-encoded across that range. The audio is copied even then: every audio frame stands on its own, so audio is cut without re-encoding at any clip length.