Closed LeonCheung closed 2 years ago
SE just uses FFmpeg from here: https://www.gyan.dev/ffmpeg/builds/
Could already be enabled... 4.x seems very fast compared to 3.x! https://docs.nvidia.com/video-technologies/video-codec-sdk/ffmpeg-with-nvidia-gpu/
Hmm, in that case, it would be very nice. Is it already enabled by default? I do see nvdec and nvenc enabled in the configurations list, but I just don't feel it, and the GPU seems not being utilized when burning it... Any hints, please? (RTX 2080ti)
I don't have an nice graphics card... but I've made a test version here: https://github.com/SubtitleEdit/subtitleedit/releases/download/3.6.4/SubtitleEditBeta.zip
How does that work?
@niksedk That's amazing, I did see the boost. Just one thing:
As tested, for a 200 secs video with ASSA, with default settings, having the "Enable CUDA hardware acceleration" disabled, it cost 25 secs to finish comparing to previous version of over 1 minute, which is really a huge boost. However, having that option enabled will cost 43 secs here... though I can see the GPU is almost fully utilized in both cases. Is that expected?
So with encoding "h264_nvenc" but with "CUDA" disabled, it took 25 secs? And with encoding "h264_nvenc" + "CUDA" enabled, it took 43 secs? And with neither "h264_nvenc" or "CUDA", it tooker a littler over 1 minute?
Exactly, tested for times and double confirmed. I'll try some longer vids tomorrow.
OK, if the longer videos are the same, I guess I'll remove the CUDA option.
Test video source info:
Setting 1: Total burning time: 06:45
Setting 2: Total burning time: 11:19
Some other thoughts:
For burning subs, a way to make full use of the hardware would be great. Not sure if it is possible...