TrainingSupportCommunity

What can be prefetched

Suggest Edits

An adaptive bit rate stream, such as HLS and DASH consists of multiple object types that are requested by a player in a specific order. You can use this behavior to prefetch these objects to speed up delivery to end users.

What's the typical flow for an HLS stream?

This discusses the object types you can usually find in an HLS stream, and the order they're presented in a request.

ObjectWhat's triggered next in a prefetchAdditional details

Master Playlist (.m3u8)

One or more variant Playlists (.m3u8).

N/A

Variant Playlist (.m3u8)

One or more segments belonging to the current variant playlist. The segment to prefetch depends on the play position of the player, and this can be difficult to determine for a VoD stream. For Live streams, this is typically one of the last three segments in the variant playlist.

For the typical HLS stream, segments have the following extensions:

  • .ts for video, video + audio
  • .aac, .ac3, .ec3 for audio-only
  • .vtt, .webvtt, .mp4, etc. for subtitles

For the newer HLS streams with CMAF segments, the extensions are:

  • .mp4, .m4v, etc. for video
  • .mp4, .m4a, etc. for audio
  • .vtt, .webvtt, .mp4 for subtitles

📘

These are not definitive, complete lists. There are a lot of supported extensions, and some are non-standard. This is to support various one-off workflows you may have in place.

Variant Playlist (.m3u8)

Zero or one init segment (described via the #EXT-X-MAP tag), that belongs to the current variant playlist.

The init segments are present almost exclusively for CMAF segments, and the typical extensions are:

  • .mp4
  • .m4v
  • .m4a

Segment

The segment that follows next in the video/audio/subtitle presentation time.

N/A

What's the typical flow for a DASH stream?

This discusses the object types you can usually find in a DASH stream, and the order they're presented in a request.

init segments for audio and video followed by actual audio and video segments. Unlike HLS, there are no bit rate-specific playlists, and content is always demuxed, so right after fetching the MPD file, the player typically requests two segments: one for video and another for audio.

ObjectWhat's triggered next in a prefetchAdditional details

Manifest File (.mpd)

N/A

Init segment

One or more content segments.

N/A

Segment

The segment that follows next in the video/audio/subtitle presentation time.

N/A

Updated 3 months ago