SVT-VP9
SVT-VP9 copied to clipboard
OpenVisualCloud
ABR without FFMPG
I'm trying to do ABR with MPEGDASH using SVT-VP9 and normally I'd be using FFMPEG with 2-pass but the encoder I'm using is Windows based and the FFMPEG plugin is not done for windows.
Are there particular settings I should use so I can get usable output for use with ABR?
Hi @jsunintel, could provide instruction to use FFMPEG plugin on Windows for svt vp9?
@japettyjohn I'm not sure that the FFMPEG plugin works for any platform, to be clear :).
@colleenkhenry That makes sense given even the dockers files explicitly omit SVT-VP9 (even though they might indicate otherwise).
FFMPEG is not a requirement - but 2-pass for ABR and thus MPEG-DASH is what I'm trying to handle.
Maybe try using bounded bitrate constant quality for now?
On Mon, May 27, 2019, 5:06 PM James Pettyjohn [email protected] wrote:
@colleenkhenry https://github.com/colleenkhenry That makes sense given even the dockers files explicitly omit SVT-VP9 (even though they might indicate otherwise).
FFMPEG is not a requirement - but 2-pass for ABR and thus MPEG-DASH is what I'm trying to handle.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/OpenVisualCloud/SVT-VP9/issues/30?email_source=notifications&email_token=AATVS2KTN4QD653QUV4SNJLPXRZO5A5CNFSM4HMSMHCKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWKUZGA#issuecomment-496323736, or mute the thread https://github.com/notifications/unsubscribe-auth/AATVS2KZCTBQIEIMWQIIJ7LPXRZO5ANCNFSM4HMSMHCA .
@japettyjohn I might be wrong here, but IIRC DASH is a live streaming format and processes video in realtime. 2 pass encoding is generally for vod and archival where a size limit is important.
Do you have a use case for 2pass+dash?
I just reviewed this and some months ago dropped the VBR 2 pass approach for CRF which does not need 2 pass. I am doing this with FFMPEG and then producing the DASH output using the Shaka packager.
2 pass is a process where the system approximates all the information and stores the compressability, motion and other parts of the video into a log during the first pass. During the second pass, it actually encodes the video. If you're not familiar with the process, I would suggest you look into it.
For streaming video this is usually not ideal since the user would need to wait for the 1st pass before even getting to see the first frame of the video. Again, I've never seen anyone use 2 for streaming, http, udp, rtmp, dash (both technically http), or otherwise.