[FFmpeg-user] Error: Impossible to convert between the formats...(Nvidia codecs)

[Dennis Mungai]

On Sun, 10 Jul 2022 at 10:28, Ben Parham <behfarteam at gmail.com> wrote:

> Hi Dennis,
> Now I am streaming with two cameras but I have a small problem, the audio
> stream is not synced with the video stream, and there is a small delay of
> around 0.2 seconds.
> How can I fix it? I would appreciate your help.
> Regards,
> Ben.
>

This somehow slipped through my email filters, seeing this now.

Try this variant and report back:

ffmpeg -y -fps_mode passthrough -hwaccel cuda -hwaccel_output_format cuda
-extra_hw_frames 3
rtsp_transport tcp -i "rtsp://admin:Pass1@192.168.1.1" -flags
+genpts -vf scale_cuda=1920:1080  -vcodec hevc_nvenc -delay 0  -preset p2
-bf:v 0 -tune hq -rc vbr -vb 700k -maxrate  800k -minrate
200k -bufsize 800k -g 25 -acodec copy -strict -2 -flags +global_header
-f rtsp "rtsp://localhost:8000/stream"

See the changes I made:

1. Completely disabled B-frames. B-frames add to encoder latency.
2. See the new option -fps_mode, which replaces the now-deprecated -vsync
parameter.
3. I also removed the +low_delay flag, as its' known to be very broken on
any input with B-frames.
4. To nvenc, I added the -delay parameter, set to 0.

The latency is propagated by the “delay” parameter (internally referred to
as async_depth) passed to FFmpeg for nvenc. The code in FFmpeg’s nvenc.c waits
until this number of extra frames have been buffered
<https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/nvenc.c#L981>
before emitting any frames and is intended to support parallel and 1:N
encoding scenarios.

The assigned default value is INT_MAX which later gets reduced down to the
number of NVENC surfaces initialized minus one, which itself has an initial
value of 4 when unset/by default.

This can be overriden by setting -delay:v 0 to fix the output latency.


Retest and report back.