ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
ecapi has quit [Ping timeout: 245 seconds]
ecapi has joined #ffmpeg
ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
Shuriko has joined #ffmpeg
durandal_1707 has quit [Ping timeout: 240 seconds]
durandal_1707 has joined #ffmpeg
FH_thecat has quit [Quit: Leaving]
lusciouslover has joined #ffmpeg
markkastoun has quit [Ping timeout: 255 seconds]
markkastoun has joined #ffmpeg
Sciencentistguy has quit [Quit: Ping timeout (120 seconds)]
Sciencentistguy has joined #ffmpeg
Dotz0cat_ is now known as Dotz0cat
lusciouslover has quit [Quit: \]
lusciouslover has joined #ffmpeg
billchenchina has quit [Read error: Connection reset by peer]
fling has quit [Ping timeout: 256 seconds]
fling has joined #ffmpeg
demon__ has quit [Read error: Connection reset by peer]
RetroPunk has quit [Quit: cya]
RetroPunk has joined #ffmpeg
haihao has quit [Ping timeout: 240 seconds]
haihao has joined #ffmpeg
friso has joined #ffmpeg
<friso>
hey guys. sometimes i start 2 encodings at the same time. then one is finished after 10 minutes while the other one has to run a little more time. the thing is when i run only 1 encoding of a video the encoding speed is around 0.56x but when i start 2 simultaniously the encoding speed is around 0.34x even when 1 of the 2 finishes before the other one the other one will get stuck on that encoding speed while it can go faster. is
<friso>
there a trick or mid-encoding command you can execute to make the gpu use its full capacity?
haihao has quit [Ping timeout: 255 seconds]
haihao has joined #ffmpeg
Muimi has quit [Quit: Going offline, see ya! (www.adiirc.com)]
<tm512>
and it seems like the issue got worse when I switched from x264 to NVENC's AVC encoder
<JEEB>
the actual error is before that
<JEEB>
since there is a reason why it decided that the input had finihsed
<JEEB>
(or the connection to your server barfed)
<JEEB>
or wherever you're pushing the ingest to
friso has quit [Quit: Leaving]
<tm512>
I'm not seeing any prior errors, though I see this message at the start of beginning the stream: [flv @ 0x560fcb97eb00] Invalid DTS: 500 PTS: 17 in output stream 0:0, replacing by guess
<tm512>
I found this which suggests not compiling with librtmp support, but then the last posts mention how that wasn't working https://trac.ffmpeg.org/ticket/7547
<tm512>
I'll still try though
<JEEB>
librtmp wouldn't change the connection getting messed up
<JEEB>
the writer is hitting an EOF which either means that the other side dropped the TCP connection, or that your input ended
kurosu has joined #ffmpeg
<tm512>
I've tried a bunch of different twitch ingest servers and this happens on every single one
<tm512>
is there any reason x11grab or NVENC would randomly cut off the input stream?
<JEEB>
I'd rather debug the output, is nvenc pushing too much data or otherwise. what is happening at the point where the other side gives up on you :P
<JEEB>
if you see no errors regarding encoding not going fast enough or anything else before that message, then it is the other side is giving up on you. welcome to network debugging
<tm512>
if librtmp couldn't be the issue I'm not sure why this issue seems to have been fixed for so many people who use the builtin RTMP instead
<JEEB>
oh you're on librtmp? I thought you meant the other way :P
<JEEB>
librtmp may be the problem but I definitely at this point on my value map see the avformat rtmp as higher than librtmp
<JEEB>
in any case, use the standard rtmp implementation in avformat and debug why the other side is giving up on you :P
<JEEB>
it's not the clear "this is your problem" answer that you probably were hoping for, but this is how these things go
<tm512>
well I'm just streaming ~3Mbit to twitch, nothing crazy
<JEEB>
I'm not saying you're doing anything crazy
<tm512>
I think I will eliminate twitch as a variable, ffmpeg can work locally as an RTMP server right?
<JEEB>
no
<JEEB>
there are projects to test things locally like nginx-rtmp
<tm512>
hmm, well I already have nginx on my home server, though it sounds like the RTMP module isn't built-in and needs to be compiled from source
<tm512>
guess I'm gonna have to learn how to load external modules into nginx
<JEEB>
yea it probably needs to. also the mediamtx project seems to be more up-to-date since it supports the recent extensions to FLV
<JEEB>
and it seems to have a docker container available
<JEEB>
(or apparently standalone binaries)
ivanich has joined #ffmpeg
<tm512>
just now I'm learning about these extensions to FLV and it's about time tbh. does ffmpeg support any of the extensions yet? hoping twitch gets on board sooner rather than later as well
lusciouslover has quit [Read error: Connection reset by peer]
Thijs has joined #ffmpeg
Thijs has quit [Client Quit]
Thijs43242342 has joined #ffmpeg
ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
<Thijs43242342>
Hi, trying to get some advice whether to submit a bug report or that it is expected behavior.
<Thijs43242342>
ffmpeg version N-112467-g5ddab49d48-20231016 Copyright (c) 2000-2023 the FFmpeg developers
<Thijs43242342>
When running two concurrent streams with this command: ffmpeg -hide_banner -threads 2 -hwaccel qsv -qsv_device /dev/dri/renderD128 -hwaccel_output_format qsv -c:v h264_qsv -avoid_negative_ts make_zero -fflags +genpts+discardcorrupt -rtsp_transport tcp -timeout 5000000 -use_wallclock_as_timestamps 1 -v 9 -loglevel 99 -i
<Thijs43242342>
One of the streams fails with the message: Assertion consumed != (-(11)) failed at libavcodec/decode.c:459
<Thijs43242342>
The commands are running in two separate LXC environments on the same device. I don't get an error when using the VAAPI device for one of the streams.
ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
realies has joined #ffmpeg
jagannatharjun has joined #ffmpeg
Volgaar has joined #ffmpeg
Muimi has quit [Quit: Going offline, see ya! (www.adiirc.com)]
<roxlu>
hi, I'm trying to resize a .mov with yuva444p12, like: `ffmpeg -i in.mov -c:v yuva444p12 -an -vf "scale=364:364" out-364x364.mov` but I get an error `yuva444p12le rawvideo cannot be written to mov`. Is this a limitation of ffmpeg as the source file contains yuva444p12? (I'm sure I'm doing something wrong)
Volgaar has quit [Ping timeout: 256 seconds]
<galad>
roxlu: what's the codec of the source file, rawvideo or something else?
AbleBacon_ has quit [Read error: Connection reset by peer]
<roxlu>
galad: prores ... ah that's probably also the issue
<roxlu>
yeah it's working now
<galad>
you didn't set any encoded, so ffmpeg tried one that can't be muxed to mov
<galad>
*encoder
<roxlu>
yeah, I didn't realise that prores was actually a codec
<galad>
oh actually you did set yuva444p12, so yeah it seems that can't be muxed to mov
<roxlu>
yeah I had to use `-c:v prores -pix_fmt yuva444p12`
ivanich has quit [Remote host closed the connection]
Thijs43242342 has joined #ffmpeg
rsx has joined #ffmpeg
Thijs43242342 has quit [Quit: Connection closed]
ivanich has joined #ffmpeg
Ogobaga has quit [Quit: Konversation terminated!]
Ogobaga has joined #ffmpeg
Gunstick- has quit [Ping timeout: 260 seconds]
Gunstick- has joined #ffmpeg
Gunstick- has quit [Read error: Connection reset by peer]
Gunstick- has joined #ffmpeg
ivanich has quit [Remote host closed the connection]
duderonomy has quit [Remote host closed the connection]
duderonomy has joined #ffmpeg
navi has joined #ffmpeg
maru_sama has joined #ffmpeg
signalhunter has quit [Quit: Zzz]
<maru_sama>
ello, I have a question regarding hardware encoding. I have an intel GPU and can use VAAPI and QSV to use hardware encoding. It works and is faster then pure software encoding. The thing is if i use handbrake with QSV encoding the speed is almos double than with ffmpeg hardware encoding. 120fps with ffmpeg compared to around 210 fps with handbrake with the same source and destination files and
<maru_sama>
locations. I tried to look at the used switches and rate controls that are used by handbrake and I am pretty sure that I am using the same with ffmpeg, but I still get this large difference.
Cindy has quit [Ping timeout: 255 seconds]
<maru_sama>
the source is h264 which is getting hardware decoded and the destination codec is hevc. Both the vaapi and qsv api give me the same performance but it is nearly half the speed I get from handbrake
whupdup has quit [Quit: Going offline, see ya! (www.adiirc.com)]
<friso>
Hey there, i've been encoding home videos from my childhood (via ffmpeg) and it's going great. There's just one problem i encounter: when i am encoding a video it usually encodes at about 5.6X speed. So every second from my VCR tape takes roughly 2 second to be encoded... Now when I ask my CPU to generate a thumbnail for the last VCR tape that has been encoded whilst another video is being encoded my video card thinks it can only
<friso>
use 0.35x processing speed for the first video encoding process while my thumbnail generation command has been completed for like forever (5 min easily). That said.. is there a nudge/command i can give to ffmpeg to re-evaulate processing capacities?
b50d_ has quit [Remote host closed the connection]
<friso>
or a command i can give my hardware graphics card to spread the word he's got more resources available?
Ogobaga has quit [Read error: Connection reset by peer]
Ogobaga has joined #ffmpeg
ivanich has quit [Remote host closed the connection]
nina_ has quit [Remote host closed the connection]
Shine_ has quit [Read error: Connection reset by peer]
ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
ecapi has quit [Remote host closed the connection]
ecapi has joined #ffmpeg
<tm512>
hmm, so I'm testing with a local RTMP server (fortunately nginx-rtmp was easy to get going for a basic test), no random disconnects so far even using librtmp
jagannatharjun has quit [Quit: Connection closed for inactivity]
ecapi has quit [Remote host closed the connection]
<tm512>
though I am just streaming an h264/aac file without reencoding, just re-muxing, so assuming this doesn't disconnect soon I think that rules out some inherent flaw in librtmp
ecapi has joined #ffmpeg
Ogobaga has quit [Quit: Konversation terminated!]
Ogobaga has joined #ffmpeg
junaid_ has quit [Ping timeout: 272 seconds]
<tm512>
still possible that it's an interaction between librtmp being fed data from the encoder, which I'll try after I'm confident that it's not going to happen without reencoding
<tm512>
if it doesn't happen locally even with reencoding then it's a twitch issue which I'd prefer it not be, definitely would rather have it be an issue with my setup since I can actually change stuff
Teukka has quit [Read error: Connection reset by peer]
Teukka has joined #ffmpeg
Teukka has quit [Changing host]
Teukka has joined #ffmpeg
dreamon has joined #ffmpeg
dreamon has quit [Client Quit]
dreamon has joined #ffmpeg
<friso>
furq, yes. but the whole process constantly takes an average of 5.6x when i'm encoding one video. and it covers 3.5x speed when i'm encoding 2 videos. so when 1 of the 2 videos is done encoding i kind of epect a signal to go from my video card or from ffmpeg to goto my video card that says "lets use all of the processing power since there is more now". i guess what you mean is that it calculates a rolling average in the
<friso>
beginning -which is fine- but it should graduately get higher. so ... i'm wondering if there is a command i can issue to my GPU to recalc or a command i can issue to ffmpeg to recalc / re-evaluate if there is more processing power. Because i'm sure there is. it just never 'rethinks'
catsoften has quit [Ping timeout: 255 seconds]
catsoften has joined #ffmpeg
catsoften has quit [Read error: Connection reset by peer]
navi has quit [Quit: WeeChat 4.0.4]
sewn has joined #ffmpeg
sewn has quit [Ping timeout: 240 seconds]
lavaball has quit [Remote host closed the connection]
catsoften has joined #ffmpeg
sewn has joined #ffmpeg
sewn has quit [Ping timeout: 272 seconds]
sewn has joined #ffmpeg
<xenorites>
Is there a good resource for the command line options needed for av1 encoding with an intel Arc GPU? I managed to get qsv av1 encoding working, but now I am running into FPS limitations - I am getting errors if my framerate is above 33, which makes me think I am missing options in my filters or the like.
luva has quit [Ping timeout: 240 seconds]
luva has joined #ffmpeg
kus has quit [Quit: Leaving]
kus has joined #ffmpeg
<BtbN>
friso: unless you specifically set stuff up to use hardware de/encoding, it's not using your graphics card at all
<BtbN>
It's technically not using your GPU itself at all
<friso>
i hear it blazing
<BtbN>
That's your CPU
<friso>
nah
<BtbN>
The video de/encode unit will not make a graphics card rev up its fans
<BtbN>
even when at full load
<friso>
i'm sure its my video card thats reving up
<BtbN>
That's not ffmpegs doing then
ecapi has quit [Remote host closed the connection]
<friso>
i think it is.. i've made it all custom
<friso>
i can drop the scripts
<friso>
but i dont think it makes a difference for you
<BtbN>
You need to specifically tell ffmpeg to use hardware acceleration
ecapi has joined #ffmpeg
<BtbN>
And even if it does, video hwaccel does not use the GPU. It's dedicated silicon
<friso>
mmkay
<BtbN>
So if your GPU feels the need to increase fan speed, something very odd is going on
<friso>
i dont know. i just figured it was all set up like this (programmed like this)
<friso>
or hardware magic was kicking in to take over processes
<BtbN>
There is no such thing, no
<friso>
but each time i start encoding i hear the fan starting up and trying to not melt my PC
<BtbN>
GPUs are very bad at video de/encoding to begin with, that's why the dedicated slicon exists
<friso>
ok
<friso>
maybe it is my CPU fan then
<BtbN>
Making nvdec and nvenc artifically load the de and encode units to 100% does not prompt my 4070 to even turn on the fan
<friso>
do you know if there is a "nudge" i can give though?
<friso>
to rethink / re-see that there is now some more capability available?
<BtbN>
But x264 encoding stuff will absolutely make my CPU run hot
<BtbN>
I'm not quire sure what you're asking
<BtbN>
if you're using hardware de/encoding, you are limited by however fast the hardware encoder can go at the given settings
<BtbN>
And with software de/encoding, your CPU speed is directly responsible for the speed
<friso>
ok. what i'm asking is:
ecapi has quit [Ping timeout: 272 seconds]
Buster__ has quit [Ping timeout: 248 seconds]
<friso>
i'm converting old home video tapes which are 45 minutes each. first i capture them 1-on-1 as 1080p video. then i encode them to 1080p h264 video files. the encoding part always, consistantly takes double as long as a 45 minute tape (1.30 hour). but sometimes i ask ffmpeg to generate a thumbnail while the encoding process is running and then all of a sudden the encoding takes way longer because the encoding speed is now NOT 0.5x but
<friso>
0.3x speed. i know for sure it can go faster but i think someone / something needs to tell FFMPEG or my hardware to try and gain more resources.
<friso>
its like it establishes the processing speed once at the beginning and then is like "hmmm.. yup this is how fast it can go"
<friso>
and i wonder if there is something i can do to make it rethink its capabilities
kurosu has quit [Quit: Connection closed for inactivity]
<tm512>
that's not how it works, unless you're doing a live capture or specifically passing in the -re flag to simulate a live capture, it will go as fast as your CPU allows
<tm512>
in those situations where the CPU is limiting things, it's normal for the encoding speed to fluctuate
ecapi has joined #ffmpeg
catsoften has quit [Ping timeout: 258 seconds]
<BtbN>
multiple processes fighting over the same resources will always be less efficient than a single process using them alone
<BtbN>
Like, what you're asking for sounds a bit like "how do I download more RAM" to me :D
<BtbN>
Your system has finite capabilities and resources. And video encoding is one of those thing that easily saturates even a very modern system