<Marth64>
For most simple "cut a segment without encoding" operations, it is more appropriate to use it before -i although both are valid scenarios. If you use it after the -i, everything up to your seek point will be decoded so your output and its timestamps will have a different result.
Tinos has quit [Ping timeout: 250 seconds]
<Marth64>
TLDR, use it before if you don't have an advanced need
<Marth64>
As you are doing a stream copy, do not expect perfect time cut but ffmpeg will do its best given the source material and where the keyframes are placed
<Marth64>
Use avidemux if you want the support of a GUI which is incredibly handy here.
hamzah has quit [Quit: Client closed]
<Marth64>
Guessing timestamps gets old after a while so it's nice to just point and click. And I do love ffmpeg but yes, GUIs are nice too :D
<noobaroo>
"If you use it after the -i, everything up to your seek point will be decoded so your output and its timestamps will have a different result"
<noobaroo>
What does this mean?
<noobaroo>
I don't understand what that means
<noobaroo>
Also, what about when encoding to AV1? Should I use it before or after the "-i" in this case?
<Marth64>
If you're encoding, just stick to using it before. It'll be frame accurate.
<Marth64>
noobaroo: is your source a DVD by chance?
jarthur has quit [Ping timeout: 268 seconds]
<noobaroo>
Nope, it's an MP4 file that was downloaded but got disconnected so the last minute of the video causes a crash
<Marth64>
Ah, sorry to hear.
<noobaroo>
I'm trying to understand the differences from those links but it's confusing, from what I can gather it doesn't matter very much, but I want to say that I don't really care about the time it takes to convert, I just want the best end product
<Marth64>
It is confusing, I don't disagree.
<aaabbb>
noobaroo: if you do "ffmpeg -ss 5:00:00 -i hugefile.mp4 ..." then it will seek instantly. if you do "ffmpeg -i hugefile.mp4 ... -ss 5:00:00 ..." then it will have to decoe (and throw away) 5 hours of content before it finally gets to do what you want
<Marth64>
Yes^
<noobaroo>
A 1 minute file is 64.9MB when used before the "-i" and 55.8MB when used after the "-i". That's a pretty big difference and although it might not make a difference to me as an end user, there is obviously a massive technical difference going on and I really wish I could understand it
<Marth64>
But with stream copy it gets funkier
<aaabbb>
noobaroo: interesting, it shouldn't be like that. are you sure both are exctly 1 minute? same exact frames?
<noobaroo>
Yup.
<noobaroo>
It seems like after the "-i" is a better choice for me
<Marth64>
noobaroo, can you ffprobe the output using -i after?
<Marth64>
What is the start time on the output?
<Marth64>
When using -ss after -i*
<Marth64>
I have observe that to screw up timestamps on streamcopy
<aaabbb>
noobaroo: /21
<aaabbb>
noobaroo: that could be because it's encoding content but "skipping" it
<aaabbb>
s/encoding/encoding or copying/
<Marth64>
Yes. Hence the funky timestamp concern
<noobaroo>
Oh wow Marth64, I'm going to upload to pastebin hold on
<noobaroo>
"What is the start time on the output?" about this
<Marth64>
It starts far ahead I assume
<noobaroo>
Massive difference I did not see until you mentioned
<noobaroo>
Yup
<Marth64>
Yes. Stick to using it before and a lot of headaches go away
<Marth64>
The couple bytes extra might be because a keyframe needed for your start point are before your start point which is ok
<Marth64>
megabytes I guess
squeaktoy has quit [Remote host closed the connection]
<Marth64>
A small difference like +-0.1 seconds is fine
<noobaroo>
Yeah
<Marth64>
Now going to my original statement, it makes more sense right?
<Marth64>
Asking because thinking of trying to phrase it better for the doc
<Marth64>
Something like this is easier to experience than explain unfortunately
<noobaroo>
I don't know how stupid I am. It's confusing to everyone I guess?
<Marth64>
I've been using this software for 2 years and -ss/-to many times a week. It's still a mystery to me how it works with streamcopy when used after -i
<Marth64>
I only "kinda" get it haha. No stupidity assumed, it's a complex software
<aaabbb>
Marth64: ffprobe with -show_frames helps clear that up
<Marth64>
Yes -show_frames is a good friend
<aaabbb>
i wrote a small utility that parses its output and gives a nice visual representation which has helped me a lot
<Marth64>
That's a good idea, aaabbb. I took the lazy route and export CSV to LibreOffice hahaha
<ppw>
I just use lossless-cut
<ppw>
it's quite brilliant, especially smart cut
<Marth64>
I always find your questions/informations here interesting aaabbb. Definitely the challenging ones :)
<Marth64>
I saw that once but couldn't stomach the JavaScript
<Marth64>
(that is not an criticism of the software or the language, I am just biased and you can ignore me now)
<aaabbb>
Marth64: this is what i use (very simple, wrote it quickly): https://bpa.st/D5ZA
<aaabbb>
but i have a more complex one that i have to clean up a bit, that also gives lots of stats
Tinos has joined #ffmpeg
<Marth64>
Thanks for sharing, added to my notes.
<aaabbb>
another i'm writing in perl which also list whether an I frame is a keyframe, size, qp, some basic stats like number of refs etc
<Marth64>
I had some b-frame nightmares a couple weeks back working on DVDs. Now that time has passed, but yeah. I feel the usefulness for sure
<aaabbb>
using both -show_frames and the trace_headers bsf
<aaabbb>
dvds are nasty wrt B frames
<Marth64>
yes. very.
<aaabbb>
iirc they don't even match the actual avc specification do they? like they don't support something that even the baseline profile literally requires
<aaabbb>
something to do with b pyramid, i don't remember
<Marth64>
DVD is MPEG-2
<furq>
i don't think anything on a dvd matches the avc speci
<furq>
yeah
<Marth64>
But, still, b-frames are nasty there... Very specifically, b-frames before timestamp 0
<aaabbb>
it's bluray i'm thinking of then lol
<aaabbb>
oof
<Marth64>
Some discs don't use them at the start, it's great. But then, there be evil discs...
Suchiman has quit [Quit: Connection closed for inactivity]
Marth64 is now known as Marth64[zzz]
Wenbin_Chen_ has joined #ffmpeg
Wenbin_Chen__ has quit [Ping timeout: 264 seconds]
Marth64 has joined #ffmpeg
Marth64[zzz] has quit [Ping timeout: 260 seconds]
twl has joined #ffmpeg
<twl>
i'm on debian unstable trying to use the nvenc hardware encoder, it's giving me the error "Driver does not support the required nvenc API version. Required: 12.1 Found: 12.0" which is indeed what version i'm on, but debian doesn't seem to give headers. i've also compiled the program myself and it says the same thing. is this an ffmpeg issue or a debian issue. https://pastebin.com/raw/4wMkD5cw
<twl>
debian doesn't seem to give headers for 12.1. i am on 12.0
Muimi has joined #ffmpeg
evilscreww has joined #ffmpeg
Marth64 has quit [Ping timeout: 268 seconds]
Marth64 has joined #ffmpeg
<MisterMinister>
Greetings, has anyone tried using ffmepg to send and recieve RIST please? Didn't find any on mailing lists, and camgoogle anything coherent with FFmpeg command samples. Transmit is working with `-f mpegts rist://225.5.5.5:25000`, but can't get receiver instance to pick up the feed... Thanks in advance.
<MisterMinister>
can't google* )
Traneptora has joined #ffmpeg
theobjectivedad has quit [Read error: Connection reset by peer]
twl has quit [Quit: WeeChat 4.1.1]
e^pi-1 has quit [Ping timeout: 268 seconds]
evilscreww has quit [Remote host closed the connection]
theobjectivedad has joined #ffmpeg
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
evilscreww has joined #ffmpeg
evilscreww has quit [Remote host closed the connection]
e^pi-1 has joined #ffmpeg
junaid_ has joined #ffmpeg
junaid_ has quit [Client Quit]
<ppw>
is there a "spek" but for video? I think I need a way to computationally determine which video is of higher quality when both appear identical
<ppw>
aaabbb: I can't frame step with this, can I?
<aaabbb>
ppw: pause then click "s" to frame step
Haripesch has joined #ffmpeg
j45 has quit [Ping timeout: 268 seconds]
j45 has joined #ffmpeg
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
j45 has quit [Ping timeout: 268 seconds]
j45 has joined #ffmpeg
evilscreww has joined #ffmpeg
<evilscreww>
It's not keeping rotation data when I convert to ts
<ppw>
aaabbb: thanks
<furq>
mpegts doesn't support rotation metadata
<evilscreww>
What the furq
<evilscreww>
You serious
ivanich has quit [Read error: Connection reset by peer]
ivanich_ has joined #ffmpeg
<evilscreww>
Not at all?
<evilscreww>
You're kidding.
<aaabbb>
evilscreww: not at all, you can either add it at the end, or maybe use the concat demuxer instead of the concat protocol (which will allow you to not stream copy to ts, idk if it drops the metadata anyway tho)
Livio has joined #ffmpeg
<evilscreww>
How does one "add it at the end"
<aaabbb>
stream copy adding rotation metadata i guess
<aaabbb>
however the new way is
vincejv has quit [Ping timeout: 260 seconds]
bitoff__ has joined #ffmpeg
<evilscreww>
I have to learn about "stream copying" vs demux copying
<aaabbb>
when you demux and mux, without any decoding and encoding, that's a stream copy
bitoff_ has quit [Ping timeout: 256 seconds]
<aaabbb>
"-c:v copy" says to stream copy, no matter what muxers/demuxers/bitstream filters you use
ivanich_ has quit [Remote host closed the connection]
ivanich has joined #ffmpeg
Some_Person has joined #ffmpeg
ivanich has quit [Read error: Connection reset by peer]
evilscreww has quit [Ping timeout: 250 seconds]
e^pi-1 has joined #ffmpeg
elastic_dog has quit [Ping timeout: 256 seconds]
e^pi-1 has quit [Quit: WeeChat 4.2.1]
Muimi has quit [Quit: Going offline, see ya! (www.adiirc.com)]
elastic_dog has joined #ffmpeg
bitblit has quit [Ping timeout: 256 seconds]
cxc has joined #ffmpeg
bitblit has joined #ffmpeg
waleee has joined #ffmpeg
MootPoot has joined #ffmpeg
e^pi-1 has joined #ffmpeg
Livio has quit [Ping timeout: 260 seconds]
beaver has joined #ffmpeg
e^pi-1 has quit [Quit: WeeChat 4.2.1]
bitoff__ has quit [Read error: Connection reset by peer]
e^pi-1 has joined #ffmpeg
bitoff__ has joined #ffmpeg
Livio has joined #ffmpeg
beaver has quit [Remote host closed the connection]
beaver has joined #ffmpeg
beaver has quit [Remote host closed the connection]
beaver has joined #ffmpeg
evilscreww has joined #ffmpeg
<evilscreww>
Do you need to make a list for concat demux or is there a way to do it manually
<aaabbb>
you need to make a list
evilscreww has quit [Ping timeout: 250 seconds]
Haripesch38 has joined #ffmpeg
Haripesch has quit [Ping timeout: 250 seconds]
psykose has quit [Remote host closed the connection]
lexano has joined #ffmpeg
low-key has quit [Remote host closed the connection]
low-key has joined #ffmpeg
low-key has quit [Remote host closed the connection]
junaid_ has quit [Ping timeout: 260 seconds]
Livio has quit [Ping timeout: 260 seconds]
yans has joined #ffmpeg
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
jemius has joined #ffmpeg
Marth64 has quit [Ping timeout: 256 seconds]
Marth64 has joined #ffmpeg
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
yans has quit [Ping timeout: 260 seconds]
junaid_ has joined #ffmpeg
rbdr has joined #ffmpeg
minimal has joined #ffmpeg
<MisterMinister>
Greetings and Salutations! Has anyone tried using FFmpeg with RIST protocol? Couldn't find any CLI samples on mailing lists and Google won't show anything coherent with FFmpeg and rist :// Transmit is running with `-f mpegts rist://225.5.5.5:25000`, but can't get receiver instance to pick up the feed...
e^pi-1 has quit [Quit: WeeChat 4.2.1]
JanC has quit [Ping timeout: 256 seconds]
JanC has joined #ffmpeg
jemius has quit [Quit: Leaving]
CAT_S is now known as CATS
lavaball has quit [Remote host closed the connection]
gchound has joined #ffmpeg
junaid_ has quit [Quit: Lost terminal]
lavaball has joined #ffmpeg
vlm has joined #ffmpeg
lavaball has quit [Remote host closed the connection]
lavaball has joined #ffmpeg
waleee has quit [Ping timeout: 252 seconds]
waleee has joined #ffmpeg
jemius has joined #ffmpeg
microchip_ has quit [Quit: There is no spoon!]
yans has joined #ffmpeg
Marth64 has quit [Ping timeout: 252 seconds]
Marth64 has joined #ffmpeg
microchip_ has joined #ffmpeg
rbdr has quit [Quit: (Zzz)]
gchound has quit [Quit: Leaving]
yans has quit [Quit: Let us play... Hide and Slay!]
waleee has quit [Ping timeout: 260 seconds]
zenstoic has joined #ffmpeg
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg
Narrat has joined #ffmpeg
iive has joined #ffmpeg
Haripesch38 is now known as Haripesch
Traneptora has quit [Quit: Quit]
rvalue has quit [Ping timeout: 268 seconds]
a0z has joined #ffmpeg
cxc has quit []
beaver has quit [Quit: Je suis normale, une personne normale]
beaver has joined #ffmpeg
rvalue has joined #ffmpeg
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg
beaver has quit [Remote host closed the connection]
beaver has joined #ffmpeg
HarshK23 has joined #ffmpeg
Suchiman has joined #ffmpeg
Tinos has joined #ffmpeg
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
Muimi has joined #ffmpeg
hamzah has joined #ffmpeg
<dv_>
does intra refresh exist for VP8? does this even make sense?
junaid_ has joined #ffmpeg
FlorianBad has quit [Remote host closed the connection]
FlorianBad has joined #ffmpeg
dreamon has joined #ffmpeg
FlorianBad has quit [Quit: Konversation terminated!]
bewilled has joined #ffmpeg
bewilled has left #ffmpeg [Leaving]
Vonter has quit [Ping timeout: 268 seconds]
Vonter has joined #ffmpeg
vincejv has quit [Remote host closed the connection]
FlorianBad has joined #ffmpeg
rv1sr has quit []
lavaball has quit [Remote host closed the connection]
lavaball has joined #ffmpeg
vincejv has joined #ffmpeg
lolok has quit [Quit: lolok]
MrZeus has joined #ffmpeg
rv1sr has joined #ffmpeg
MrZeus has quit [Read error: Connection reset by peer]
MrZeus has joined #ffmpeg
vincejv has quit [Quit: Bye bye! Leaving for now...]
jemius has quit [Quit: Leaving]
MrZeus has quit [Read error: Connection reset by peer]
MrZeus has joined #ffmpeg
MrZeus has quit [Read error: Connection reset by peer]
MrZeus has joined #ffmpeg
vincejv has joined #ffmpeg
waleee has joined #ffmpeg
<klaxa>
JEEB: happy to report that the guy that came here the other day for ffserver sent patches and they fixed things. so ffserver can now be operated the way it used to (still not mature :P)
<klaxa>
went ahead and switched to the new avfifo api and libmicrohttpd api in the process as well
lusciouslover has quit [Quit: \]
lusciouslover has joined #ffmpeg
rv1sr has quit []
jafa has joined #ffmpeg
<jafa>
quick one - how to I tell ffmpeg which video decoder to use? when transcoding to a different video format?
<jafa>
the option -c:v xxx sets the encoder codec to use
<jafa>
I am testing qsv and it reports mpeg2video (mpeg2_qsv) -> h264 (h264_qsv)
<jafa>
I want to run some tests doing software decode -> qsv encode
<jafa>
ie use mpeg2video, not mpeg2_qsv
<BtbN>
the -c option sets either the de or the encoder
<BtbN>
depending on whether it's specififed for an input or output
<BtbN>
But pretty sure for mpeg2 it's all the same decoder, and just a normal hwaccel
<jafa>
ah, that worked
<jafa>
doing mpeg2video (mpeg2_qsv) -> h264 (h264_qsv) is leaking memory badly... I an trying to isolate it to the decode or the encode
minimal has quit [Quit: Leaving]
junaid_ has quit [Quit: Lost terminal]
<klaxa>
ffmpeg rust rewrite when? :^)
<JEEB>
and then you find out that your memory usage is in the hwaccel framework
<JEEB>
jafa: you could try running it under valgrind's massif
<JEEB>
there's even a Qt-based massif visualizer
<jafa>
trying AMD encode to compare but hit a different problem...
<jafa>
it reports mpeg2video (native) -> h264 (h264_vaapi) which seems right
<jafa>
Impossible to convert between the formats supported by the filter 'Parsed_null_0' and the filter 'auto_scale_0'
<jafa>
JEEB: will take a look at valgrind shortly - thanks for the idea
theobjectivedad has quit [Ping timeout: 260 seconds]
<jafa>
ah, need to add hwupload to the filter chain
theobjectivedad has joined #ffmpeg
tips__ has joined #ffmpeg
tips_ has quit [Ping timeout: 272 seconds]
<jafa>
is there something similar to intel_gpu_top for amd?
<BtbN>
the Windows task manager
<jafa>
using Linux
<jafa>
sw decode + amd encode doesn't leak memory
zenstoic has quit [Quit: Connection closed for inactivity]
linext has joined #ffmpeg
<linext>
what's the command to remove video between two times?
<jafa>
sw decode + intel encode leaks memory
rbdr has joined #ffmpeg
beaver has quit [Quit: Lost terminal]
<BtbN>
if by "intel encode" you mean qsv... yeah. I wouldn't bother with that on Linux
<jafa>
yes - qsv
<BtbN>
Just use vaapi
<jafa>
can I configure the intel deinterlacer vi vaapi?
<jafa>
s/vi/via
<BtbN>
I mean, there's a deinterlacing filter for vaapi
Haripesch has quit [Ping timeout: 250 seconds]
rbdr has quit [Client Quit]
<jafa>
the intel gpu has a very good deinterlacer... so I am using the intel gpu for interlaced content. can use intel or amd for progressive content
<jafa>
good as in better than yadif
rbdr has joined #ffmpeg
<jafa>
specifying vaapi and pointing to the intel gpu triggers h264_qsv codec and it is leaking memory
sonicrules1234 has quit [Ping timeout: 260 seconds]
FlorianBad has quit [Quit: Konversation terminated!]
sonicrules1234 has joined #ffmpeg
Keshl has quit [Ping timeout: 268 seconds]
Tinos has quit [Remote host closed the connection]
FlorianBad has joined #ffmpeg
Keshl has joined #ffmpeg
lavaball has quit [Remote host closed the connection]
a0z has quit [Ping timeout: 260 seconds]
MrZeus has quit [Read error: Connection reset by peer]
MrZeus has joined #ffmpeg
GenTooMan has quit [Ping timeout: 272 seconds]
<jafa>
running valgrind
Livio has joined #ffmpeg
Narrat has quit [Quit: They say a little knowledge is a dangerous thing, but it's not one half so bad as a lot of ignorance.]
<jafa>
weird - valgrind is reporting "driftup" over and over as expected but no real leaks on exit (just 4k, mostly strdup strings)
<jafa>
the leak is bad enough that ffmpeg keeps being killed by the linux oom killer
<jafa>
which is amusing because the machine has 64GB RAM
<noobaroo>
I have a TV that refuses to preserve aspect ratio, it always stretches my 1920x816 video to fill the whole screen. Is there a way to artificially add black bars on the top and bottom of the video to make it artificially 1920x1080 so my TV will play it correctly?
Marth64 has quit [Ping timeout: 260 seconds]
Marth64 has joined #ffmpeg
Tinos has joined #ffmpeg
vlm has quit [Quit: Leaving]
<jafa>
does ffmpeg track memory enough to free it on exit when a codec is leaking memory? odd the valgrind can see the memory usage growing but no leak on exit
<BtbN>
You can easily just accumulate tons of memory without leaking it
<jafa>
ffmpeg memory usage when doing h264_qsv encode grows and grosw until linux kills the process
<jafa>
valgrind reports "in use at exit: 4,249 bytes in 9 blocks" (ie nothing), and "total heap usage: 2,614,159 allocs, 2,614,150 frees, 4,826,343,493 bytes allocated" (this was after a few minutes of runtime)