-
Notifications
You must be signed in to change notification settings - Fork 160
Chunk Method
Av1an splits the source video into small chunks that are encoded in parallel. There are several different ways to split a video file, each with pros and cons. This option is set by --chunk_method CHUNK_METHOD
.
Chunk Method | Frame accurate? | No intermediate files? | Fast? | No other dependencies? |
---|---|---|---|---|
select | ✔️ | ✔️ | ❌ | ✔️ |
vs_ffms2 | ✔️ | ✔️ | ✔️ | ❌ |
hybrid | ✔️ | ❌ | ✔️[1] | ✔️ |
[1] hybrid is slightly slower
- For aom reuse first pass (
--reuse_first_pass
), the chunk method must be frame accurate.
This method uses ffmpeg's select filter to only output the frames in a given chunk from the source file. This method needs to decode all frames that occur before a chunk, so for chunks near the end of a long source, this takes a long time. This method can take as much as 30% longer to complete an encode than segment on a 30 minute 30 fps video.
This method uses VapourSynth and FFMS2 to perform fast seeks to frames. Both of these must be installed and configured to use this method.
This method combines segment and select together. If a scene cut happens to have a keyframe in the source file, segment is used for that chunk. If not, a small file containing multiple chunks is generated with segment. Each individual chunk is accessed from this smaller file with select. This allows for frame accurate seeking, while limiting the number of frames that select must decode. This method takes a little longer than segment and vs_ffms2, but is no where near as slow as just select.