|
2024-05-16, 03:25 AM
I had an idea that could potentially speed up the generation of Trickplay images in a significant way:
https://features.jellyfin.org/posts/2672...lay-images I'd love to know your thoughts on the feasibility of this method and the effort required to implement it as an option. 
1
2024-05-16, 12:41 PM
(This post was last modified: 2024-05-16, 12:42 PM by TheDreadPirate.)
My understanding is that the current process can use the GPU for decoding and CPU for encoding. Part/most of the reason it takes so long is that the entire file needs to be decoded. Based on some other observations, I believe that is a limitation of ffmpeg and not of Jellyfin.
Jellyfin 10.10.0 (Docker)
Ubuntu 24.04 LTS Intel i3 12100 Intel Arc A380 OS drive - SK Hynix P41 1TB Storage 3x WD Red Pro 6TB CMR in RAIDZ1 (JF Library) ![[Image: GitHub%20Sponsors-grey?logo=github]](https://img.shields.io/badge/GitHub%20Sponsors-grey?logo=github){kind=icon}
2024-05-16, 01:26 PM
We have an alternate approach that is even faster pending for 10.10: https://github.com/jellyfin/jellyfin/pull/11511
This will make an hour video's trickplay images to be generated in 5-6 seconds, but the thumbnail will be less precise and some hardware video decoder is not compatible with this(so you have to use CPU in these cases) 
1
|
|
|
|
|
