Jesper

Do you mean on the thumbnail that is created?

Okay, i will add this in a future release!

I can look into getting more confidence values in the binary_sensor. It would also be possible to make the sensitivty configurable. Im thinking of implementing a more sophisticated version of...

yeah that is definitely not optimal... I find it hard to find good examples on the web on how to perform highly accurate face recognition sadly

Will do, seems easy enough! Basically just need to allow it in the config schema since ffmpeg supports it

You dont need to specify hwaccel args when using the `*_nvenc` decoders/encoders, try removing that. If that doesnt work please paste your full config so i can look at it.

You are right, its `h264_nvenc`/`hevc_nvenc` for encoding and `h264_cuvid`/`hevc_cuvid` for decoders

Hmm are you sure your cameras RTSP stream is `hevc`? Can you show me the output of `docker exec -it viseron ffprobe rtsp://user:[email protected]:554/cam/realmonitor?channel=1&subtype=1` You need to add your credentials to...

Okay that is interesting! Can you play around a little with the command and see if its one of the default flags that are causing issues? Start with a really...

Yes that seems very likely. Does this work for you or does it also error out? ``` docker run -it --rm --entrypoint "" roflcoopter/amd64-cuda-viseron:1.8.0 ffmpeg -hide_banner -loglevel error -avoid_negative_ts make_zero...