Recherche avancée

Sur d’autres sites (4597)

• memory leak on FFMPEG RTSP streams ?

  1er août 2024, par user3260912

  I have been having an issue recently where I have several IP cameras set up and running. Every so often, one of the streams will lock the entire virtual machine it is running in up. Looking at top, Ffmpeg is using 86% of 64 GB of RAM when the machine locks up. Yes, I realize my IP camera password


  


  As you can see from the screenshot, the RTSP stream process is using 51 GB of RAM to pull the images to a jpeg file.

  


  Any ideas if this is a known issue ?

  


  ffmpeg version 4.4.2-0ubuntu0.22.04.1+esm4 Copyright (c) 2000-2021 the FFmpeg developers

  


• FFmpeg stream extraction modifies subtitles [closed]

  21 mai 2024, par user18812922

  I have a video with the following ffprobe output :

  


  Input #0, matroska,webm, from 'video.mkv':
  Metadata:
    title           : Video - 01
    creation_time   : 2021-07-14T02:49:59.000000Z
    ENCODER         : Lavf58.29.100
  Duration: 00:22:57.28, start: 0.000000, bitrate: 392 kb/s
  Chapters:
    Chapter #0:0: start 0.000000, end 86.169000
      Metadata:
        title           : Opening
    Chapter #0:1: start 86.169000, end 641.266000
      Metadata:
        title           : Part A
    Chapter #0:2: start 641.266000, end 651.359000
      Metadata:
        title           : Eyecatch
    Chapter #0:3: start 651.359000, end 1286.160000
      Metadata:
        title           : Part B
    Chapter #0:4: start 1286.160000, end 1356.355000
      Metadata:
        title           : Ending
    Chapter #0:5: start 1356.355000, end 1376.876000
      Metadata:
        title           : Preview
  Stream #0:0: Video: hevc (Main 10), yuv420p10le(tv, bt709), 854x480 [SAR 1280:1281 DAR 16:9], 23.98 fps, 23.98 tbr, 1k tbn (default)
      Metadata:
        DURATION        : 00:22:56.959000000
  Stream #0:1(eng): Audio: vorbis, 48000 Hz, stereo, fltp (default)
      Metadata:
        title           : English [FLAC 2.0]
        DURATION        : 00:22:57.278000000
  Stream #0:2(jpn): Audio: vorbis, 48000 Hz, stereo, fltp
      Metadata:
        title           : Japanese [FLAC 2.0]
        DURATION        : 00:22:57.276000000
  Stream #0:3(eng): Subtitle: ass (ssa)
      Metadata:
        title           : Signs and Songs [FMA1394/Redc4t]
        DURATION        : 00:22:51.090000000
  Stream #0:4(eng): Subtitle: ass (ssa)
      Metadata:
        title           : English [FMA1394/Redc4t]
        DURATION        : 00:22:51.090000000
  Stream #0:5(eng): Subtitle: hdmv_pgs_subtitle (pgssub), 1920x1080
      Metadata:
        title           : Full English Retail
        DURATION        : 00:22:51.120000000
  Stream #0:6: Attachment: ttf
      Metadata:
        filename        : 8bitoperator.ttf
        mimetype        : application/x-truetype-font
  Stream #0:7: Attachment: ttf
      Metadata:
        filename        : Cabin-Bold.ttf
        mimetype        : application/x-truetype-font
  Stream #0:8: Attachment: ttf
      Metadata:
        filename        : calibrib.ttf
        mimetype        : application/x-truetype-font
  Stream #0:9: Attachment: ttf
      Metadata:
        filename        : daniel_0.ttf
        mimetype        : application/x-truetype-font
  Stream #0:10: Attachment: ttf
      Metadata:
        filename        : DEATH_FONT.TTF
        mimetype        : application/x-truetype-font
  Stream #0:11: Attachment: ttf
      Metadata:
        filename        : Dominican.ttf
        mimetype        : application/x-truetype-font
  Stream #0:12: Attachment: ttf
      Metadata:
        filename        : gishabd.ttf
        mimetype        : application/x-truetype-font
  Stream #0:13: Attachment: ttf
      Metadata:
        filename        : PATRICK_0.TTF
        mimetype        : application/x-truetype-font
  Stream #0:14: Attachment: ttf
      Metadata:
        filename        : Qlassik-Medium.ttf
        mimetype        : application/x-truetype-font
Unsupported codec with id 98304 for input stream 6
Unsupported codec with id 98304 for input stream 7
Unsupported codec with id 98304 for input stream 8
Unsupported codec with id 98304 for input stream 9
Unsupported codec with id 98304 for input stream 10
Unsupported codec with id 98304 for input stream 11
Unsupported codec with id 98304 for input stream 12
Unsupported codec with id 98304 for input stream 13
Unsupported codec with id 98304 for input stream 14


  


  I am trying to extract the subtitles, edit them and reattach them to the video.
(I need my program to do that so I don't want to use other software)

  


  Command 1

  


  ffmpeg -i video.mkv -map 0:3 -c:s ssa subs.ass
ffmpeg -i video.mkv -i subs.ass -map 0 -map -0:s -map 1 -c copy out.mkv


  


  Command 2

  


  ffmpeg -i video.mkv -map 0:3 subs.ass
ffmpeg -i video.mkv -i subs.ass -map 0 -map -0:s -map 1 -c copy out.mkv


  


  Command 3

  


  ffmpeg -i video.mkv -map 0:3 subs.srt
ffmpeg -i video.mkv -i subs.srt -map 0 -map -0:s -map 1 -c copy out.mkv


  


  Command 4

  


  ffmpeg -i video.mkv -map 0:3 subs.srt
ffmpeg -i subs.srt subs.ass
ffmpeg -i video.mkv -i subs.ass -map 0 -map -0:s -map 1 -c copy out.mkv


  


  Command 5

  


  ffmpeg -i video.mkv -map 0:3 subs.ass
ffmpeg -i subs.ass subs.srt
ffmpeg -i video.mkv -i subs.srt -map 0 -map -0:s -map 1 -c copy out.mkv


  


  The problem

  


  After extraction the subtitles seem to be really quick, meaning they are displayed and disappear really quickly.

  


  For example the first subtitle is as follows in srt :

  


  1&#xA;00:00:03,100 --> 00:00:03,560&#xA;<font face="Dominican" size="77" color="#f7f7f7">Within the spreading darkness</font>&#xA;

  


  Now, in srt it also has wrong size but I assume that's because of the conversion from ass to srt.

  


  If I reattach the subtitle file in the video and open it, it is displayed and disappears way too fast and it doesn't match the original subtitles in the video.

  


  (ie, the original video subtitles are showing for at least a second)

  


  Expected behaviour

  


  The subtitles should be displayed for the same duration as the original subtitles.

  


  NOTE

  


  It's my first question for ffmpeg related issues so feel free to ask me for anything else you may need.

  


  UPDATE 1

  


  I realized that the subtitles were ok for the timings as they had the same line multiple times, so the problem for not playing is something else.

  


  Example of the file

  


  1&#xA;00:00:03,100 --> 00:00:03,560&#xA;<font face="Dominican" size="77" color="#f7f7f7">Within the spreading darkness</font>&#xA;&#xA;2&#xA;00:00:03,560 --> 00:00:04,650&#xA;<font face="Dominican" size="77" color="#f7f7f7">Within the spreading darkness</font>&#xA;&#xA;3&#xA;00:00:04,650 --> 00:00:05,100&#xA;<font face="Dominican" size="77" color="#f7f7f7">Within the spreading darkness</font>&#xA;

  


  So the problem is that VLC doesn't show more than the first subtitle.

  


  The strange thing is when I use the below command

  


  ffmpeg -i video.mkv -i subs.srt -map 0 -map -0:s -map 1 -c copy -c:s subrip out.mkv


  


  Then more lines of the subtitle (but not all) play.

  


  It stops at the 17th line.

  


  I believe that's an encoder's problem ? but I really don't know.

  


  Also what I noticed is that VLC stops the subtitles but Windows Media Player (Windows 11 version) display the subtitles correctly even after the 17th line.

  


  BUT, if I add subtitles from another video they are played correctly in both VLC and Windows Media Player.

  


  Update 2
As @Gyan said in his answer I should use the following command

  


  ffmpeg -i video.mkv -map 0:3 -c:s copy subs.ass


  


  But then if I attach the subs again with

  


  ffmpeg -i video.mkv -i subs.ass -map 0 -map -0:s -map 1 -c copy -c:s ass out.mkv


  


  The subtitles show up to 17th line in both VLC and Windows Media Player.

  


  or

  


  ffmpeg -i video.mkv -i .\subs.ass -map 0 -map -0:s -map 1 -c copy out.mkv


  


  The subtitles do not show up at all. (Not even in Windows Media Player)

  


• Consent Mode v2 : Everything You Need to Know

  7 mai 2024, par Alex — Analytics Tips

  Confused about Consent Mode v2 and its impact on your website analytics ? You’re not the only one. 

  Google’s latest update has left many scratching their heads about data privacy and tracking. 

  In this blog, we’re getting straight to the point. We’ll break down what Consent Mode v2 is, how it works, and the impact it has.

  What is Consent Mode ?

  What exaclty is Google Consent Mode and why is there so much buzz surrounding it ? This question has been frustrating analysts and marketers worldwide since the beginning of this year. 

  Consent Mode is the solution from Google designed to manage data collection on websites in accordance with user privacy requirements.

  This mode enables website owners to customise how Google tags respond to users’ consent status for cookie usage. At its core, Consent Mode adheres to privacy regulations such as GDPR in Europe and CCPA in California, without significant loss of analytical data.

  

  How does Consent Mode work ?

  Consent Mode operates by adjusting the behaviour of tags on a website depending on whether consent for cookie usage is provided or not. If a user does not consent to the use of analytical or advertising cookies, Google tags automatically switch to collecting a limited amount of data, ensuring privacy compliance.

  This approach allows for continued valuable insights into website traffic and user behavior, even if users opt out of most tracking cookies.

  What types of consent are available in Consent Mode ?

  As of 6 March 2024, Consent Mode v2 has become the current standard (and in terms of utilising Google Advertising Services, practically mandatory), indicating the incorporation of four consent types :

  1. ad_storage : allows for the collection and storage of data necessary for delivering personalised ads based on user actions.
  2. ad_user_data : pertains to the collection and usage of data that can be associated with the user for ad customisation and optimisation.
  3. ad_personalization : permits the use of user data for ad personalisation and providing more relevant content.
  4. analytics_storage : relates to the collection and storage of data for analytics, enabling websites to analyse user behaviour and enhance user experience.

  Additionally, in Consent Mode v2, there are two modes :

  1. Basic Consent Mode : in which Google tags are not used for personalised advertising and measurements if consent is not obtained.
  2. Advanced Consent Mode : allows Google tags to utilise anonymised data for personalised advertising campaigns and measurements, even if consent is not obtained.

  What is Consent Mode v2 ? (And how does it differ from Consent Mode v1 ?)

  Consent Mode v2 is an improved version of the original Consent Mode, offering enhanced customisation capabilities and better compliance with privacy requirements. 

  The new version introduces additional consent configuration parameters, allowing for even more precise control over which data is collected and how it’s used. The key difference between Consent Mode v2 and Consent Mode v1 lies in more granular consent management, making this tool even more flexible and powerful in safeguarding personal data.

  In Consent Mode v2, the existing markers (ad_storage and analytics_storage) are accompanied by two new markers :

  1. ad_user_data – does the user agree to their personal data being utilized for advertising purposes ?
  2. ad_personalization – does the user agree to their data being employed for remarketing ?

  In contrast to ad_storage and analytics_storage, these markers don’t directly affect how the tags operate on the site itself. 

  They serve as additional directives sent alongside the pings to Google services, indicating how user data can be utilised for advertising purposes.

  While ad_storage and analytics_storage serve as upstream qualifiers for data (determining which identifiers are sent with the pings), ad_user_data and ad_personalization serve as downstream instructions for Google services regarding data processing.

  How is the implementation of Consent Mode v2 going ?

  The implementation of Consent Mode v2 is encountering some issues and bugs (as expected). The most important thing to understand :

  1. Advanced Consent Mode v2 is essential if you have traffic and campaigns with Google Ads in the European Union.
  2. If you don’t have substantially large traffic, enabling Advanced Consent Mode v2 will likely result in a traffic drop in GA4 – because this version of consent mode (unlike the basic one) applies behavioural modelling to users who haven’t accepted the use of cookies. And modelling the behaviour requires time.

  The aspect of behavioural modelling in Consent Mode v2 implies the following : the data of users who have declined tracking options begin to be modelled using machine learning. 

  However, training the model requires a suitable data volume. As the Google’s documentation states :

  The property should collect at least 1,000 events per day with analytics_storage=’denied’ for at least 7 days. The property should have at least 1,000 daily users submitting events with analytics_storage=’granted’ for at least 7 of the previous 28 days.

  Largely due to this, the market’s response to the Consent Mode v2 implementation was mixed : many reported a significant drop in traffic in their GA4 and Google Ads reports upon enabling the Advanced mode. Essentially, a portion of the data was lost because Google’s models lacked enough data for training. 

  And from the very beginning of implementation, users regularly report about a few examples of that scenario. If your website doesn’t have enough traffic for behaviour modelling, after Consent Mode v2 switching you will face significant drop in your traffic in Google Ads and GA4 reports. There are a lot of cases of observing 90-95% drop in metrics of users and sessions.

  In a nutshell, you should be prepared for significant data losses if you are planning to switch to Google Consent Mode v2.

  How does Consent Mode v2 impact web analytics ? 

  The transition to Consent Mode v2 alters the methods of user data collection and processing. The main concerns arise from the potential loss of accuracy and completeness of analytical data due to restrictions on the use of cookies and other identifiers when user consent is absent. 

  With Google Consent Mode v2, the data of visitors who have not agreed to tracking will be modelled and may not accurately reflect your actual visitors’ behaviours and actions. So as an analyst or marketer, you will not have true insights into these visitors and the data acquired will be more generalised and less accurate.

  Google Consent Mode v2 appears to be a kind of compromise band-aid solution. 

  It tries to solve these issues by using data modelling and anonymised data collection. However, it’s critical to note that there are specific limitations inherent to the modelling mechanism.

  This complicates the analysis of visitor behavior, advertising campaigns, and website optimisation, ultimately impacting decision-making and resulting in poor website performance and marketing outcomes.

  Wrap up

  Consent Mode v2 is a mechanism of managing Google tag operations based on user consent settings. 

  It’s mandatory if you’re using Google’s advertising services, and optional (at least for Advanced mode) if you don’t advertise on Google Ads. 

  There are particular indications that this technology is unreliable from a GDPR perspective. 

  Using Google Consent Mode will inevitably lead to data losses and inaccuracies in its analysis. 

  In other words, it in some sense jeopardises your business.