Extract poster frames from local camera videos
This commit is contained in:
@@ -133,15 +133,18 @@ in the overview.
|
||||
and a moment to start. Playback uses the Vita's hardware decoder via
|
||||
SceAvPlayer, so only MP4 (H.264/AAC) plays — Immich's transcoded
|
||||
`video/playback` stream is used, which is H.264 with default server
|
||||
settings.
|
||||
settings. The decoder rotates four output frame buffers (~13 MB CDRAM at
|
||||
1080p) so async GPU rendering can't be torn by a recycled frame.
|
||||
- Caps at 1000 server assets and 500 local camera files per launch; no
|
||||
albums or search.
|
||||
- Local media sync is one-way (Vita → server). It scans `ux0:picture` and
|
||||
`ux0:video/CAMERA` by default (the standard camera locations), recursing
|
||||
~2 levels; use `syncdir=` in the config to scan other folders.
|
||||
- Local videos have no poster-frame thumbnail (the Vita has no still
|
||||
decoder for arbitrary MP4 frames), so their grid cell is a dark
|
||||
placeholder with the VIDEO badge; opening one still plays it.
|
||||
- Local video grid cells show a poster thumbnail (with the VIDEO badge):
|
||||
the first frame is decoded with SceAvPlayer on the background loader
|
||||
thread (NV12 -> RGB, downscaled), the same handoff as photos. Poster
|
||||
extraction and playback share the single hardware decoder, so a poster is
|
||||
never decoded while a video is playing (and vice versa).
|
||||
- Hashing 500 files at startup takes a while; it runs in the background so
|
||||
browsing is never blocked, but statuses fill in gradually.
|
||||
- `deviceId` is hard-coded to "PS Vita" and uploads are not associated with
|
||||
|
||||
Reference in New Issue
Block a user