Closed limjoe closed 6 months ago
ffmpeg \
-hwaccel cuda -i input.ts -vf "subtitles=subtitles.srt:force_style='Alignment=2,MarginV=20',hwupload_cuda" \
-c:v h264_nvenc -profile:v main -preset medium -rc:v vbr -cq:v 19 -bf 0 \
-acodec aac \
-copyts
-y output.ts
You are now able to setup the video codec parameters for transcript.
Strongly recommand setting the -bt 0
to be compatible with both live stream and WebRTC.
Fixed in v5.14.11
What is the business background? Please provide a description. AI transcript task is slow, especially ffmpeg encode part, so the transcript live stream with subtitle delay a lot.
Is your feature request related to a problem? Please describe. Large delay and poor user experience
Describe the solution you'd like AI transcript support GPU for ffmpeg encode, for example NVIDIA GPU for accelerate.
Describe alternatives you've considered
Additional context