15:50 <EmberCrest> Hey folks, I have a question. I'm using MediaMTX to stream to ffmpeg. However, after a recent update of both programs, I couldn't use rtsp basic auth anymore. MediaMTX sent back a 401 Unauthorized.

15:50 <EmberCrest> (in response to ANNOUNCE)

15:50 <EmberCrest> Checked the code and found that MediaMTX is expecting an "Authorization" header. Is there a way to force FFMPEG to send this header?

15:51 <EmberCrest> Because I don't see it in the request.

17:37 <flo> Hey, can someone help me with getting the proper command for my use case?

17:43 <microchip_> no

17:45 <flo> Cool, thanks

17:46 <another|> just ask

17:50 <flo> Here's the output I get: https://pastebin.com/PmvZAus0 I need to make sure that all frames stay at exactly the same count as in the (obviously partly broken) input file, to make sure the output is still in sync to additional audio sources. I tried with chatgpt, but it seems like with the command it told me after much discussion, the broken frames are still dropped, but the file has the same

17:50 <flo> length overall. I don't care what happens with the broken frames, I just need to make sure all working frames stay at their position

19:06 <johnjaye> which filter is best for dropping frames in a time interval?

19:06 <johnjaye> i think i can adapt the zoompan filter to insert extra frames. but i want to delete them without splitting and reattaching the videos

19:58 <BtbN> fps?

19:58 <BtbN> like, if you want to drop every second frame, halve the FPS

20:03 <johnjaye> no i mean a discrete time period like from 5.6 to 7 seconds

20:03 <johnjaye> i thought it would be called something simple like erase or drop but don't see one

20:08 <johnjaye> and i don't think -ss or -t will do it

20:41 <BtbN> select then

22:55 <aaabbb> is there a simple and authoritative source that i can give to someone who stubbornly believes (and without testing) that hardware-accelerated video encoding provides the same quality per bitrate as software encoding?

22:56 <BtbN> just try it and see

22:57 <aaabbb> they still think it's just as good and they don't want to use ssim to test it

22:57 <BtbN> Just test more extreme scenarios

22:58 <BtbN> of course if you give both plenty of bitrate or a chessmatch to encode, it'll look fine enough

22:59 <aaabbb> perhaps the fact that he has to increase bitrate by 4x when the pixel count increases by 4x might be a reason? does nvenc use tiling or something, because that's a very huge increase in bitrate for only 4x increase in pixels

23:00 <BtbN> That's not how that works, what?

23:00 <aaabbb> i know

23:00 <aaabbb> but that's what he says he has to do. said the same thing about going from 30 to 60fps...

23:00 <aaabbb> so i wonder if nvenc might be so bad that he actually *does* have to do that?

23:01 <BtbN> modern nvenc is pretty good

23:01 <BtbN> on an old card it's pretty bad

23:01 <BtbN> but it's never linear anyway

23:01 <aaabbb> yeah it's much better than it used to be. but 4x?

23:03 <aaabbb> maybe it is using tiling and that's why he claim he has to increase the bitrate by 4x to increase pixel count by 4x

23:05 <aaabbb> anyway, i'll keep looking for a simple and authoritative source, since i know he won't be willing to actually test more extreme scenarios (or use ssim etc)