Speed drops after extended clip-sequence export
MilkAndBanana1 opened this issue · comments
Bug description
Used it auto-editor to a batch file.
Using hevc_nvenc. (ryzen 5 5600H,3060 laptop gpu)
Inputted a 27 minute mkv file.
First 50-70 files are fast, after 100+ it starts to slow down considerably.
What's your platform?
Windows.
What command did you use
set specialparams=--my-ffmpeg --video-codec hevc_nvenc --edit audio:stream=0,mincut=25,minclip=3 --margin 0.05sec,0.2sec --export clip-sequence
rem ##################### SPLITTING THE CLIPS
md edited_clips
set "folder=%~dp0\edited_clips"
set /p inputFile=Enter the name of the file to open:
auto-editor.exe "%inputFile%" -o "edited_clips/clips.mp4" !specialparams!
What properties does your input video have?
.\podcast23.mkv:
- video:
- track 0:
- codec: h264
- fps: 19001/317
- resolution: 1920x1080
- aspect ratio: 16:9
- pixel aspect ratio: 1:1
- pix fmt: yuv420p
- color range: tv
- color space: bt709
- color primaries: bt709
- color transfer: bt709
- timebase: 1/1000
- track 0:
- audio:
- track 0:
- codec: aac
- samplerate: 48000
- channels: 2
- track 1:
- codec: aac
- samplerate: 48000
- channels: 2
- track 0:
- container:
- duration: 1629.044
- bitrate: 102590866
Metadata:
ENCODER : Lavf60.3.100
Duration: 00:27:09.04, start: 0.000000, bitrate: 102590 kb/s
Stream #0:0: Video: h264 (High), yuv420p(tv, bt709, progressive), 1920x1080 [SAR 1:1 DAR 16:9], 59.94 fps, 59.94 tbr, 1k tbn
Metadata:
DURATION : 00:27:09.044000000
Stream #0:1: Audio: aac (LC), 48000 Hz, stereo, fltp
Metadata:
title : Track2
DURATION : 00:27:08.992000000
Stream #0:2: Audio: aac (LC), 48000 Hz, stereo, fltp
Metadata:
title : Track3
DURATION : 00:27:08.992000000
Comments
Have a feeling this is just CPU getting fried after multiple encodes but issuing this in case there's a potential optimization.
Update: Merged the audio tracks and ran it through ffmpeg before doing so and it seems to be faster.
I already know auto-editor can be slow when exporting as a clip-sequence when there are hundreds of clips. it's an O(n^2) operation due to file seeking being linear. I don't consider this a valid bug because I only have so much time on this earth to be spending time optimizing every last use case.