November 2024 - ffmpeg-user - lists.ffmpeg.org

Creating MP4 with G711 audio
by d.gordenin＠ngslab.ru 22 Nov '24

22 Nov '24

Hello. I'm trying to make a mp4 file from a video/audio stream. When audio is AAC it works fine. But I cannot pack G711U audio. |av_register_all(); avcodec_register_all(); av_log_set_callback(nullptr); av_format_context_ = avformat_alloc_context(); if (!av_format_context_) { std::string error_msg = "Failed to allocate avformat context"; log_->Error(error_msg); throw std::runtime_error(error_msg.c_str()); } av_format_context_->oformat = av_guess_format("mp4", filename.c_str(), NULL); strcpy(av_format_context_->filename, filename.c_str()); AVCodec* codec = avcodec_find_encoder(AV_CODEC_ID_PCM_MULAW); av_format_context_->oformat->audio_codec = codec->id; audio_stream_ = avformat_new_stream(av_format_context_, codec); if (audio_stream_ == nullptr) { log_->Error("Can not create audiostream."); return false; } AVCodecContext& codec_context = *audio_stream_->codec; if (avcodec_copy_context(audio_stream_->codec, av_format_context_->streams[1]->codec) != 0) { log_->Error("Failed to Copy Context"); return false; } audio_stream_->time_base = { 1, codec_context.sample_rate }; if (av_format_context_->oformat->flags & AVFMT_GLOBALHEADER) codec_context.flags |= AV_CODEC_FLAG_GLOBAL_HEADER; log_->Info("Init output {}", filename); av_dump_format(av_format_context_, 0, filename.c_str(), 1); int ret = avio_open(&av_format_context_->pb, filename.c_str(), AVIO_FLAG_WRITE); if (ret < 0) { log_->Error("Can not open file for writing."); return false; } ret = avformat_write_header(av_format_context_, NULL); if (ret < 0) { log_->Error("Can not write header."); return false; } | Here avformat_write_header() returns -22. What's wrong here? Which data should I supply to this function?

4 4

strange pts in v4l2 capture
by Casey Connor 18 Nov '24

18 Nov '24

Hi - I am capturing to an .avi file in linux, and the resulting file has stuttering audio. If I demux the streams (1 video and 1 stereo audio) they play fine on their own. If I simply remux the broken file like this: ffmpeg -i test.avi -c:a copy -c:v copy test-remuxed.avi ...the "remuxed" version plays fine. If I compare the pts in the broken vs. remuxed file using ashowinfo, I notice something: the audio pts values don't match the nb_samples -- instead they are 4x larger. E.g. for the first line see how 1024 nb_samples causes the next pts to be 4096, and for the next line nb_samples:256 increases pts to 5120, and so on: Broken file: [Parsed_ashowinfo_0 @ 0x70654c002f80] n:0 pts:0 pts_time:0 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:1024 checksum:73F22CAC plane_checksums: [ 73F22CAC ] [Parsed_ashowinfo_0 @ 0x70654c002f80] n:1 pts:4096 pts_time:0.0853333 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:256 checksum:929680FC plane_checksums: [ 929680FC ] [Parsed_ashowinfo_0 @ 0x70654c002f80] n:2 pts:5120 pts_time:0.106667 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:1024 checksum:7A55BBD5 plane_checksums: [ 7A55BBD5 ] With the working remuxed file, the next pts is simply the previous pts + nb_samples: [Parsed_ashowinfo_0 @ 0x76a788002f80] n:0 pts:0 pts_time:0 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:1024 checksum:73F22CAC plane_checksums: [ 73F22CAC ] [Parsed_ashowinfo_0 @ 0x76a788002f80] n:1 pts:1024 pts_time:0.0213333 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:256 checksum:929680FC plane_checksums: [ 929680FC ] [Parsed_ashowinfo_0 @ 0x76a788002f80] n:2 pts:1280 pts_time:0.0266667 fmt:s16 channels:2 chlayout:stereo rate:48000 nb_samples:1024 checksum:7A55BBD5 plane_checksums: [ 7A55BBD5 ] I'm using ffmpeg version 7.0.2-3ubuntu1 on Kubuntu 24.10. I'm capturing 720x480 NTSC_M from a Hauppauge USB-Live2 S/Video capture card (cs231xx driver) like this: ffmpeg -f alsa -ac 2 -i hw:CARD=Cx231xxAudio,DEV=0 -f v4l2 -i /dev/video2 -c:a copy -c:v rawvideo test.avi Also note: when capturing I get numerous errors like these: [vost#0:0/rawvideo @ 0x55f451f6b280] Clipping frame in rate conversion by 0.011955 [aost#0:1/copy @ 0x55f451f75440] Non-monotonic DTS; previous: 29662, current: 29564; changing to 29663. This may result in incorrect timestamps in the output file. Thanks in advance for any ideas! -c

1 0

sidechain compression: NOT ducking, but envelope
by Michael Werzowa 18 Nov '24

18 Nov '24

Dear List, I am new to the list. Searched the last two years of the list for „sidechain“, and did not find a hint. My usecase: Using an audio file to provide an envelope, that is applied to another audio file, modulating the gain of this file. With ffmpeg I easily can manage „ducking“, via direct call and using the python interface, but with my use case I run into errors or get the wrong output. For music mixing, this would be a typical use of sidecahin compression. ChatGPT 4o and Claude 3.5 Sonnet gave me a lot of advice, which all lead to the same: either the wrong effect, or an error. An example that does NOT work, although it should be the thing I want: ffmpeg -i data/to_be_modulated.wav -i data/sidechain.wav -filter_complex "[1:a]aformat=sample_fmts=fltp:sample_rates=44100:channel_layouts=stereo,volume=2,highpass=f=20,agate=mode=downward:threshold=0:ratio=1:attack=0.0057:release=0.02:range=60:makeup=1[env];[0:a][env]amultiply" data/output-processed.wav Ducking worked with the same files. I tried different parameter changes, but I am lost. Does anyone of you have a solution for my use case? Is there a „tutorial“ for using the -fiter_complex? With librosa and numpy I managed to do this, but this is much slower. I would need „nearly realtime“, hopefully about 2sec. My librosa/numpy solution for this takes about 10sec on a MacBook Pro Max M1. Thank you for your effort, best regards, Michael -- ******************************************************************* Michael Werzowa A-1190 Wien, Siolygasse 20/3B/6 +43 664 302 4511, fax +43 810 9554 185931 mikey(a)werzowa.at *******************************************************************

3 4

Cannot seek an HLS stream with epoch timecode
by George Daramouskas 17 Nov '24

17 Nov '24

Hi everyone, I have an srt stream running 24/7, at some point in time I'd like to start an HLS stream that has as its input the srt stream. Now, the HLS stream could go down for arbitrary reasons and I am fine with that, it should be able to be restarted and keep going, however because I will be logging some events that take place at a point in time, I need to be able to refer to that point in time regardless of when the HLS stream started. This has led me to using unix epoch as a starting offset, specifically: ` ffmpeg -re -i srt://$url/?latency=20&mode=caller <http://10.21.6.126:9001/?latency=20&mode=caller> -map 0:v -map 0:a -hls_list_size 0 -master_pl_name stream1.m3u8 -strftime 1 -hls_segment_filename /core/data/stream1/%Y%m%d/segment-%s.ts -strftime_mkdir 1 -output_ts_offset $(date +%s) -hls_flags program_date_time /core/data/stream1.m3u8 ` Using ffprobe, I am able to verify that pts time is correct and gets incremented correctly: ` ffprobe -show_packets $url ........ [PACKET] codec_type=audio stream_index=1 pts=83020451563104 pts_time=1729592740.898000 dts=83020451563104 dts_time=1729592740.898000 duration=N/A duration_time=N/A size=344 pos=9420493 flags=K_ [/PACKET] ..... ` pts_time and dts_time look correct, same for the rest of the a/v streams. Now I want to seek on that hls and cut chunks out of it: ` ./ffmpeg_g -y -ss 1729592740.898000 -t 2 -seek_timestamp 1 -i https://10.21.6.110/stream1.m3u8 /tmp/test123.mp4 ` I am getting `could not seek to position 1729592740.898` and I end up with an improper mp4 cut from live instead from the right timestamp. Am I misusing the cli somehow here? Looking forward to hearing your thoughts Kind regards, George Daramouskas

2 1

Re: [FFmpeg-user] Questions about `-to` parameter
by Vincent Deconinck 15 Nov '24

15 Nov '24

> > I look up the ffmpeg document You have probably missed the part of https://trac.ffmpeg.org/wiki/Seeking that says 'Note when "-ss" used before "-i" only: ("-ss" as input option) the timestamps will be reset according to 0.͏ So "-t" and "-to" shall be the same.' :-) [Q1] Is this really intentional? If it is, can anyone share the thought of > this design? > There is a logic, at least. Think "-ss before -i" as two separate ffmpeg operations: 1 : create a new file without the start 2 : truncate the new file at a certain time Operation 2 has no idea what the original file was, as an "input seeking" (step1) was applied before, so it doesn't know how much was removed. > [Q2] Is there a way to make the second command works like the first > command? > Not that I know of. You should just compute the desired duration (as end-start) before calling ffmpeg, and I suggest you use -t instead of -to in that case. They behave the same and that removes any ambiguity for the reader. Vincent

2 1

Questions about `-to` parameter
by Pan Lee 15 Nov '24

15 Nov '24

Hi, I am using ffmpeg to crop video based on this command: ffmpeg -i input.mp4 -ss 23:43 -to 40:00 -c:v copy -c:a copy output.mp4 but this is a little bit slow, so I try to switch the -ss parameter before -i, which looks like: ffmpeg -ss 23:43 -i input.mp4 -to 40:00 -c:v copy -c:a copy output.mp4 when I execute this command, the result is not what I expected. The duration of output video is 40 minutes, not 17 minutes. I look up the ffmpeg document and ask the ChatGPT, and figure out that the timestamp of -to parameter is a relative stop time of -ss timestamp, just like the -t parameter. And the ChatGPT says this is intentional. [Q1] Is this really intentional? If it is, can anyone share the thought of this design? [Q2] Is there a way to make the second command works like the first command? All the best, lp

1 0

RETRIEVE FRAME RATE WITH DECIMAL VALUE
by Hervé ANSELME 14 Nov '24

14 Nov '24

Hi, using ffprobe release 6.1, I can retrieve the frame rate of a film with the following command : ffprobe -v error -select_streams v:0 -show_entries stream=avg_frame_rate -of default=noprint_wrappers=1 "the video.mkv" which for example, retrieves the frame rate value in fractional form : 24000/1001 I wonder if there is a way to retrieve the value, in decimal form, rather than fractional form ; for example : 23,976215 If anybody knows a solution, I'd appreciate. Thanks in advance for your answers and advices. Regards, HA [https://s-install.avcdn.net/ipm/preview/icons/icon-envelope-tick-round-oran…]<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campai…> Sans virus.www.avast.com<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campai…>

3 3

Technical Problem
by amazonprincessdiana＠gmail.com 12 Nov '24

12 Nov '24

Hello, I usually use 4K Video Downloader to download videos from YouTube channels I subscribe to. A couple of the channels I keep an eye on have a really bad practice of giving their videos really bad names that are extremely long and contain really stupid characters in them. The crazy naming they give the videos has been causing routine problems with 4K not completing the final step of the downloading process which is compiling the videos leaving them as failed downloads that are in the format of .part files. I have tried use ffmpeg to do that final merging and have even worked with ChatGPT to have it create a batch file that I can keep in each channel's folder to run to do the process and it keeps failing to convert the videos. I spent several hours trying to get ChatGPT to correct the problem but nothing is working. I use ChatGPT because I never learned any sort of scripting or command line usage over the 25+ years I have been in IT because I am too used to GUI and Windows. I am on a 1 year old self built water cooled desktop PC with excellent hardware and software and I am on Windows 11 Pro. Do you have any suggestions or can you improve ffmpeg so that it reliably works even with files with long names with crazy characters? I am willing to share any information you need in the way of error messages and even some sample files. I have zero connection to the creators so I cannot tell them to "get a brain" and name their videos with better names. Thank you. Sincerely AmazonPrincessDiana

5 5

Trying to understand ffmpeg cpu usage in this case of capturing rtsp streams
by Jim Ruxton 10 Nov '24

10 Nov '24

I'm using Ubuntu 22.04 and ffmpeg version 4.4.2-0 I'm capturing 3 rtsp streams as shown at the bottom . Stream A resolution is 1600x1200 Stream B and Stream C resolution is 384x192 . Stream A and B are coming from one device (ip address) and Stream C is coming from a second device (ip address) When I start Stream A and capture it with the command below I get a cpu usage of around 90% using the top command. If I then start stream B simultaneously it shows about 25% cpu usage in another ffmpeg instance. Then with the other 2 streams running I start Stream C and it shows about 90% cpu usage. If I start Stream B on it's own it shows 90% cpu usage. In other words it doesn't seem to matter what the resolution of the video coming in is but rather which stream starts first that determines the cpu usage. Does it matter that Stream A an B are coming from the same ip address and that explains why the second stream uses less cpu? My expectation is that Stream B and C would have the same cpu usage since they are the same resolution just coming from different devices. Can anyone explain what I see as strange behaviour. Stream commands follow. Stream A: ffmpeg -f rtsp -i rtsp://192.168.35.243:8086/?camera=world -f v4l2 /dev/video2 Stream B: ffmpeg -f rtsp -i rtsp://192.168.35.243:8086/?camera=eyes -vf format=yuv420p -f v4l2 /dev/video3 Stream C: ffmpeg -f rtsp -i rtsp://192.168.35.247:8086/?camera=eyes -vf format=yuv420p -f v4l2 /dev/video5 Thanks, Jim

5 13

Record multiple channel audio from Focusrite Scarlett 18i8
by Stuart Robinson 06 Nov '24

06 Nov '24

As in the title I am wondering if it is possible to record all inputs from my focusrite to separate mono tracks. ffmpeg -sources dshow just shows [analogue 1 + 2 (audio)] despite the device having 18 inputs (I don't care about the spdif and AES inputs but I would like to have the main input channels) Does anyone know if it is possible to record these channels with ffmpeg? Do I need to make a change to dshow somehow? Thanks, Stuart. -- Thanks, Stuart Robinson, Senior Technician, uCreate Makerspace, Main Library, University of Edinburgh, 30 George Square, Edinburgh, EH8 9LD 0131 651 5001 The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.

2 4