Recherche avancée
Autres articles (96)
-
MediaSPIP 0.1 Beta version
25 avril 2011, parMediaSPIP 0.1 beta is the first version of MediaSPIP proclaimed as "usable".
The zip file provided here only contains the sources of MediaSPIP in its standalone version.
To get a working installation, you must manually install all-software dependencies on the server.
If you want to use this archive for an installation in "farm mode", you will also need to proceed to other manual (...) -
HTML5 audio and video support
13 avril 2011, parMediaSPIP uses HTML5 video and audio tags to play multimedia files, taking advantage of the latest W3C innovations supported by modern browsers.
The MediaSPIP player used has been created specifically for MediaSPIP and can be easily adapted to fit in with a specific theme.
For older browsers the Flowplayer flash fallback is used.
MediaSPIP allows for media playback on major mobile platforms with the above (...) -
ANNEXE : Les plugins utilisés spécifiquement pour la ferme
5 mars 2010, parLe site central/maître de la ferme a besoin d’utiliser plusieurs plugins supplémentaires vis à vis des canaux pour son bon fonctionnement. le plugin Gestion de la mutualisation ; le plugin inscription3 pour gérer les inscriptions et les demandes de création d’instance de mutualisation dès l’inscription des utilisateurs ; le plugin verifier qui fournit une API de vérification des champs (utilisé par inscription3) ; le plugin champs extras v2 nécessité par inscription3 (...)
Sur d’autres sites (9646)
-
CRO Program : Best Practices and KPIs to Track [2024]
8 mai 2024, par ErinDriving traffic to your website is only one part of the equation ; the second part is getting those visitors to convert by completing a desired action — creating an account, signing up for a newsletter or completing a purchase.
But if you fail to optimise your website for conversions, you’ll have a hard time guiding visitors further down the funnel and turning them into customers.
That’s where a CRO program (or conversion rate optimisation) can help.
This article will cover conversion rate optimisation best practices and outline key metrics and KPIs to start tracking to see an improvement in your conversion rates.
What is a CRO program ?
In the simplest terms, a CRO program — also called a CRO plan — is a digital marketing strategy. It focuses on implementing different tactics that can lead to an increase in conversion rate and maximising revenue.
One thing to remember is that the definition of “conversion” varies from business to business. The most obvious type of conversion would be a financial transaction or a completed form — but it comes down to what you consider a valuable action.
Many different actions can count as conversions, depending on your marketing goals.
Besides making a purchase, other common examples of key conversion moments include creating a new account, signing up for a free trial, booking a demo and subscribing to an email newsletter.
Another thing worth noting is that while the average conversion rate on e-commerce websites is 3.76%, it might fluctuate across different industries and device types. Case in point — desktop devices have higher conversion rates than mobile devices, clocking in at 4.79% and 3.32%, respectively.
So, in addition to defining your key conversion moments, you should also go over conversion insights relevant to your specific industry.
The importance of conversion rate optimisation
You’d be right to assume that the ultimate goal of a conversion rate optimisation process is to drive revenue through higher conversion rates — but don’t focus solely on the numbers. The core principle of a CRO program is improving the customer experience. Once you’ve achieved that, the increase in conversion rate will follow.
According to a recent report, global conversion rate optimisation (CRO) software sales are expected to reach $3.7 billion by 2032 — up from $1.1 billion in 2021.
This growth indicates the increasing interest in strategies and tools that can help optimise the conversion funnel. Businesses are looking for ways to keep potential customers engaged and improve the average conversion rate — without necessarily increasing their spending.
Here are a few reasons why a CRO program deserves a spot in your broader digital marketing strategies :
- It can lower your cost per acquisition (CPA) : A CRO program is about optimising your conversion funnel by leveraging existing assets and website traffic rather than increasing your spending — which lowers the costs of acquiring new customers and, in turn, drives ROI.
- It can maximise customer lifetime value (CLV) : If you can turn one-time buyers into repeat customers, you’ll be one step closer to building a loyal user base and increasing your CLV.
- It can lead to increased sales and boost your revenue : Higher conversion rates typically mean higher revenue ; that’s arguably the most obvious benefit of implementing a CRO program.
- It improves the overall user experience : The goal is to make your site more accessible, easier to navigate and more engaging. Delivering the experience people want — and expect — when navigating your website is one of the core principles of a CRO program.
- It helps you to get to know your customers better : You can’t meet your customers’ needs without taking the time to know them, create user personas and understand their preferences, pain points and conversion barriers they may be facing.
Conversion optimisation gives you a competitive edge in revenue and brand reputation.
5 CRO best practices
Here are five conversion rate optimisation strategies and best practices that can make a real difference in the customer experience — and drive potential conversions.
Create a CRO roadmap in advance
First and foremost, you’ll need a well-defined “game plan” that aligns with and reflects your conversion goals.
A CRO roadmap is a detailed manual that outlines how to implement different elements of your CRO-related efforts. Marketing teams can refer to this step-by-step framework for test planning, prioritisation and resource allocation while optimising their marketing strategy.
While conversion rate optimisation can be a complex process — especially when you don’t know what to tackle first — we’ve found that there are three things you need to consider when setting the foundations of a successful CRO program :
- The “why” behind your website traffic : You’re likely using different online marketing strategies — from SEO to pay-per-click (PPC). So, it’s best to start by gathering channel-specific conversion insights through marketing attribution. Then identify which of these efforts have the biggest impact on your target audience.
- The so-called “conversion blockers” that tell you where and why visitors tend to leave without completing a desired action : Funnel analysis might reveal problematic pages — drop-off points where you tend to lose most of your visitors.
- Your “hooks” : User feedback can be of great help here ; you can learn a lot by simply asking your customers to fill out a quick online survey and tell you what motivated them to take action.
Before working on that “game plan,” perform a pre-test analysis.
Matomo combines web analytics and user behaviour analytics with features like Heatmaps, Session Recordings, Form Analytics, Funnel Analytics, A/B Testing and User Flow. It can give you those initial benchmarks for measuring progress and a potential increase in conversion rate.
Validate your ideas with A/B and multivariate testing
Conversion rate optimisation is an iterative process. So, it shouldn’t come as a surprise that A/B testing variants of page layouts, CTAs, headlines, copy and other elements is a big part of it.
Multivariate and A/B testing allows you to test a wide range of elements across your site and identify what works — and, more importantly, what doesn’t — in terms of driving conversions.
On that note, Matomo’s A/B Testing feature can support your conversion rate optimisation process by identifying variants that perform better based on statistical significance.
Try Matomo for Free
Get the web insights you need, without compromising data accuracy.
Get to know your website visitors
Driving conversions comes down to understanding potential customer’s pain points and needs — and delivering an experience that positions you as the solution and gets them to take action.
Here are a few things that can help you understand your website visitors better :
- Collecting customer feedback through surveys and using it to identify main areas for improvement
- Creating detailed customer personas and optimising your website design and messaging based on your target audience’s pain points, needs and wants
- Using heatmaps — colour-coded data visualisation tools that illustrate user interactions — and scroll maps to get a comprehensive overview of online sessions and identify the most engaging elements and those that stand out as potential conversion barriers
Matomo’s Heatmaps can help you identify the most-clicked elements on the page and show how far users scroll — providing powerful user insights you can use to optimise these pages.
Try Matomo for Free
Get the web insights you need, without compromising data accuracy.
Remove friction points
As we previously discussed, identifying friction points and barriers to conversion — issues that prevent visitors from converting — is one of the crucial aspects of developing a CRO plan.
Many different “conversion blockers” are worth looking into, including :
- Lengthy or otherwise complex checkout processes
- No guest checkout feature
- Device type, browser and OS compatibility issues
- Slow site speed and other technical issues
- Lack of free shipping and limited payment methods
- Absence of social proof (customer reviews and testimonials) and trust badges
Once you’ve identified what’s slowing down or completely discouraging users from reaching key conversion moments, take the time to address it.
Switch to text-based CTAs
Calls-to-action (CTAs) play a crucial role in guiding customers from interest to action. However, sometimes they fail to do their job — encouraging website visitors to proceed to the next step — effectively.
The most obvious reason is that your CTAs aren’t visually engaging or clear enough. In that case, you can try using action-oriented language and stronger visual elements and aligning the CTA copy with the context of the page.
But more often than not, the issue comes down to a phenomenon called “banner blindness” — the tendency of website visitors to ignore (either intentionally or unintentionally) elements on a page that resemble banner ads.
And if that’s what’s preventing visitors from converting, consider switching to text-based CTAs.
Conversion rate optimisation metrics and KPIs
At this point, you should know the outcomes you hope to achieve. Your next step should be to figure out how you’re going to measure and analyse results — and identify the changes that made the most impact on your conversion funnel.
After all, your CRO action plan should be based on data — assumptions and “gut feelings” will rarely lead to a notable increase in conversion rates.
That brings us to key performance indicators (KPIs) :
Tracking CRO metrics and website KPIs can help you understand the customer’s journey and path to purchase, identify opportunities for improving the user experience (UX) and determine how to optimise conversions.
That said, you shouldn’t try to track every metric in the book ; think about your ultimate goal and identify the metrics and KPIs most relevant to your business.
We’ll assume that you’re already tracking macro- and micro-conversions. However, we’ve outlined a few additional key conversion rate optimisation metrics you should keep an eye on to make sure that your CRO program is performing as intended :
- Cost-per-conversion : By measuring how much you spend on each successful conversion — again, completed forms, sign-ups and sales all count as key conversion moments — you’ll be in a better position to assess the cost-effectiveness of your online marketing strategies.
- Starter rate : This metric tells you the number of people who start filling out the form, after seeing it. This metric is particularly important for companies that rely on getting leads from forms.
- Average order value (AOV) : This metric is important for e-commerce sites to understand the value of their transactions. AOV calculates the average monetary value of each order.
That’s not all ; you can also use a web analytics tool like Matomo to gain granular insights into visitors :
- Unique, new and returning visitors : Tracking the number of new and returning visitors your website gets within a given timeframe will help you understand your user base and determine if your content resonates with them. While you want a constant stream of new traffic, don’t overlook the importance of returning visitors ; they’re the foundation of a loyal customer base.
- User flows : By analysing the user flows, you’ll have a visual representation of how visitors use your website, which will help you understand their journey and the specific path they take.
- Bounce rate : This metric tells you how many users viewed a single page on your site and ended up leaving before they took any kind of action. As such, it’s a clear indicator of how good your content, CTAs and website layout are at keeping users engaged.
- Exit rate : Another key metric to track is the exit rate — the percentage of users who drop off at a specific page. High-exit pages usually lack important information and CTAs, cause frustration or otherwise fail to meet users’ expectations. Keep in mind that there’s a difference between bounce rate and exit rate — the latter involves users who viewed at least one other page.
There are many other user engagement metrics you should keep an eye on in addition to the ones mentioned above — including time on-page, actions per visit, scroll depth and traffic source. You’ll find all this information — and more — in Matomo’s Page Analytics Report.
Conclusion
Implementing a CRO program can be a time-consuming and iterative process. However, it’s vital for guiding your marketing efforts and making data-driven decisions that’ll ultimately help you drive growth and reach your business goals.
It’s best to start by identifying where your website visitors come from and what contributes to — or prevents them from — taking further action. But that’s easier said than done. You’ll need to leverage web analytics tools like Matomo to gather powerful user insights and monitor your website’s performance.
As an all-in-one, privacy-friendly web analytics solution, Matomo combines traditional web analytics and advanced behavioural analytics — delivering a consistent experience based on 100% accurate, unsampled data.
Join the 1 million websites that have chosen Matomo as their web analytics platform. Start your 21-day free trial today — and see how Matomo can help you improve your website’s conversion rates. No credit card required.
Try Matomo for Free
21 day free trial. No credit card required.
-
I want to take any Audio from a file and encode it as PCM_ALAW. My Example is a .m4a file to .wav file
22 novembre 2023, par ClockmanI have been working on this for a while now while am generally new to ffmpeg library, I have studied it a bit. The challenge I have that at the point of witting to file I get the following exception.


"Exception thrown at 0x00007FFACA8305B3 (avformat-60.dll) in FfmpegPractice.exe : 0xC0000005 : Access violation writing location 0x0000000000000000.". I understand this means am writing to an uninitialized buffer am unable to discover why this is happening. The exception call stack shows the following


avformat-60.dll!avformat_write_header() C
avformat-60.dll!ff_write_chained() C
avformat-60.dll!ff_write_chained() C
avformat-60.dll!av_write_frame() C
FfmpegPractice.exe!main() Line 215 C++


Some things I have tried


This code is part of a larger project built with CMake but for some reason I could no step into ffmpeg library while debugging, So I recompiled ffmpeg ensured debugging was enabled so I could drill down to the root cause but I still could not step into the ffmpeg library.


I then created a minimal project using visual studio c++ console project and I still could not step into the code.


I have read through many ffmpeg docs and some I could find on the internet and I still could not solve it.


This is the code


#include <iostream>

extern "C" {
#include <libavcodec></libavcodec>avcodec.h>
#include <libavformat></libavformat>avformat.h>
#include <libswresample></libswresample>swresample.h>
#include <libavutil></libavutil>opt.h>
#include <libavutil></libavutil>audio_fifo.h>
}

using namespace std;

//in audio file
string filename{ "rapid_caller_test.m4a" };
AVFormatContext* pFormatCtx{};
AVCodecContext* pCodecCtx{};
AVStream* pStream{};

//out audio file
string outFilename{ "output.wav" };
AVFormatContext* pOutFormatCtx{ nullptr };
AVCodecContext* pOutCodecCtx{ nullptr };
AVIOContext* pOutIoContext{ nullptr };
const AVCodec* pOutCodec{ nullptr };
AVStream* pOutStream{ nullptr };
const int OUTPUT_CHANNELS = 1;
const int SAMPLE_RATE = 8000;
const int OUT_BIT_RATE = 64000;
uint8_t** convertedSamplesBuffer{ nullptr };
int64_t dstNmbrSamples{ 0 };
int dstLineSize{ 0 };
static int64_t pts{ 0 };

//conversion context;
SwrContext* swr{};

uint32_t i{ 0 };
int audiostream{ -1 };


void cleanUp() 
{
 avcodec_free_context(&pOutCodecCtx);;
 avio_closep(&(pOutFormatCtx)->pb);
 avformat_free_context(pOutFormatCtx);
 pOutFormatCtx = nullptr;
}

int main()
{

/*
* section to setup input file
*/
if (avformat_open_input(&pFormatCtx, filename.data(), nullptr, nullptr) != 0) {
 cout << "could not open file " << filename << endl;
 return -1;
}
if (avformat_find_stream_info(pFormatCtx, nullptr) < 0) {
 cout << "Could not retrieve stream information from file " << filename << endl;
 return -1;
}
av_dump_format(pFormatCtx, 0, filename.c_str(), 0);

for (i = 0; i < pFormatCtx->nb_streams; i++) {
 if (pFormatCtx->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
 audiostream = i;
 break;
 }
}
if (audiostream == -1) {
 cout << "did not find audio stream" << endl;
 return -1;
}

pStream = pFormatCtx->streams[audiostream];
const AVCodec* pCodec{ avcodec_find_decoder(pStream->codecpar->codec_id) };
pCodecCtx = avcodec_alloc_context3(pCodec);
avcodec_parameters_to_context(pCodecCtx, pStream->codecpar);
if (avcodec_open2(pCodecCtx, pCodec, nullptr)) {
 cout << "could not open codec" << endl;
 return -1;
}

/*
* section to set up output file which is a G711 audio
*/
if (avio_open(&pOutIoContext, outFilename.data(), AVIO_FLAG_WRITE)) {
 cout << "could not open out put file" << endl;
 return -1;
}
if (!(pOutFormatCtx = avformat_alloc_context())) {
 cout << "could not create format conext" << endl;
 cleanUp();
 return -1;
}
pOutFormatCtx->pb = pOutIoContext;
if (!(pOutFormatCtx->oformat = av_guess_format(nullptr, outFilename.data(), nullptr))) {
 cout << "could not find output file format" << endl;
 cleanUp();
 return -1;
}
if (!(pOutFormatCtx->url = av_strdup(outFilename.data()))) {
 cout << "could not allocate file name" << endl;
 cleanUp();
 return -1;
}
if (!(pOutCodec = avcodec_find_encoder(AV_CODEC_ID_PCM_ALAW))) {
 cout << "codec not found" << endl;
 cleanUp();
 return -1;
}
if (!(pOutStream = avformat_new_stream(pOutFormatCtx, nullptr))) {
 cout << "could not create new stream" << endl;
 cleanUp();
 return -1;
}
if (!(pOutCodecCtx = avcodec_alloc_context3(pOutCodec))) {
 cout << "could not allocate codec context" << endl;
 return -1;
}
av_channel_layout_default(&pOutCodecCtx->ch_layout, OUTPUT_CHANNELS);
pOutCodecCtx->sample_rate = SAMPLE_RATE;
pOutCodecCtx->sample_fmt = pOutCodec->sample_fmts[0];
pOutCodecCtx->bit_rate = OUT_BIT_RATE;

//setting sample rate for the container
pOutStream->time_base.den = SAMPLE_RATE;
pOutStream->time_base.num = 1;
if (pOutFormatCtx->oformat->flags & AVFMT_GLOBALHEADER)
 pOutCodecCtx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;

if (avcodec_open2(pOutCodecCtx, pOutCodec, nullptr)) {
 cout << "could not open output codec" << endl;
 cleanUp();
 return -1;
}
if ((avcodec_parameters_from_context(pOutStream->codecpar, pOutCodecCtx)) < 0) {
 cout << "could not initialize stream parameters" << endl;
} 

AVPacket* packet = av_packet_alloc();

swr = swr_alloc();
swr_alloc_set_opts2(&swr, &pOutCodecCtx->ch_layout, pOutCodecCtx->sample_fmt, pOutCodecCtx->sample_rate,&pCodecCtx->ch_layout, pCodecCtx->sample_fmt, pCodecCtx->sample_rate, 0, nullptr);
swr_init(swr);

int ret{};
int bSize{};
while (av_read_frame(pFormatCtx, packet) >= 0) {
 AVFrame* pFrame = av_frame_alloc();
 AVFrame* pOutFrame = av_frame_alloc();
 if (packet->stream_index == audiostream) {
 ret = avcodec_send_packet(pCodecCtx, packet);
 while (ret >= 0) {
 ret = avcodec_receive_frame(pCodecCtx, pFrame);
 if (ret == AVERROR(EAGAIN))
 continue;
 else if (ret == AVERROR_EOF)
 break;
 dstNmbrSamples = av_rescale_rnd(swr_get_delay(swr, pCodecCtx->sample_rate) + pFrame->nb_samples, pOutCodecCtx->sample_rate, pCodecCtx->sample_rate, AV_ROUND_UP);
 if ((av_samples_alloc_array_and_samples(&convertedSamplesBuffer, &dstLineSize, pOutCodecCtx->ch_layout.nb_channels,dstNmbrSamples, pOutCodecCtx->sample_fmt, 0)) < 0) {
 cout << "coult not allocate samples array and buffer" << endl;
 }
 int channel_samples_count{ 0 };
 channel_samples_count = swr_convert(swr, convertedSamplesBuffer, dstNmbrSamples, (const uint8_t**)pFrame->data, pFrame->nb_samples);
 bSize = av_samples_get_buffer_size(&dstLineSize, pOutCodecCtx->ch_layout.nb_channels, channel_samples_count, pOutCodecCtx->sample_fmt, 0);
 cout << "no of samples is " << channel_samples_count << " the buffer size " << bSize << endl;
 pOutFrame->nb_samples = channel_samples_count;
 av_channel_layout_copy(&pOutFrame->ch_layout, &pOutCodecCtx->ch_layout);
 pOutFrame->format = pOutCodecCtx->sample_fmt;
 pOutFrame->sample_rate = pOutCodecCtx->sample_rate;
 if ((av_frame_get_buffer(pOutFrame, 0)) < 0) {
 cout << "could not allocate output frame samples " << endl;
 av_frame_free(&pOutFrame);
 }
 
 //populate out frame buffer
 av_frame_make_writable(pOutFrame);
 for (int i{ 0 }; i < bSize; i++) {
 pOutFrame->data[0][i] = convertedSamplesBuffer[0][i];
 cout << pOutFrame->data[0][i];
 }
 if (pOutFrame) {
 pOutFrame->pts = pts;
 pts += pOutFrame->nb_samples;
 }
 int res = avcodec_send_frame(pOutCodecCtx, pOutFrame);
 if (res < 0) {
 cout << "error sending frame to encoder" << endl;
 cleanUp();
 return -1;
 }
 //int er = avformat_write_header(pOutFormatCtx,nullptr);
 AVPacket* pOutPacket = av_packet_alloc();
 pOutPacket->time_base.num = 1;
 pOutPacket->time_base.den = 8000;
 if (pOutPacket == nullptr) {
 cout << "unable to allocate packet" << endl;
 }
 while (res >= 0) {
 res = avcodec_receive_packet(pOutCodecCtx, pOutPacket);
 if (res == AVERROR(EAGAIN))
 continue;
 else if (ret == AVERROR_EOF)
 break;
 av_packet_rescale_ts(pOutPacket, pOutCodecCtx->time_base, pOutFormatCtx->streams[0]->time_base);
 //av_dump_format(pOutFormatCtx, 0, outFilename.c_str(), 1);
 if (av_write_frame(pOutFormatCtx, pOutPacket) < 0) {
 cout << "could not write frame" << endl;
 }
 }
 }
}
 av_frame_free(&pFrame);
 av_frame_free(&pOutFrame);
}
if (av_write_trailer(pOutFormatCtx) < 0) {
 cout << "could not write file trailer" << endl;
}
swr_free(&swr);
avcodec_free_context(&pOutCodecCtx);
av_packet_free(&packet);
}
</iostream>

Error/Exception


The exception is thrown when I call


if (av_write_frame(pOutFormatCtx, pOutPacket) < 0) { cout << "could not write frame" << endl; }
I also called this line

//int er = avformat_write_header(pOutFormatCtx,nullptr);

to see if I will get an exception but it did not throw any exception.


I have spent weeks on this issue with no success.
My goal is to take any audio from a file an be able to resample it if need be, and transcode it to PCM_ALAW.
I will appreciate any help I can get.


-
SegFault while freeing nvenc hwdevice_ctx
23 février 2024, par camelCaseFor a project, I created a class encoding the output of an OpenGL
renderbufferobject usingh264_nvenc. Unfortunately, tidying up doesn't work, and the program crashes with a SegFault. The reason is accessing an unaccessable memory region, happening twice in the final lines (see below) when callingav_buffer_unref( &_hwDeviceRefCtx )and implicitely also duringavcodec_free_context( &_pCodecCtx ), but both calls are required for shutting down.

The (in this case relevant)
valgrind-output is

Invalid read of size 8
 at 0x48AD987: UnknownInlinedFun (buffer.c:121)
 by 0x48AD987: UnknownInlinedFun (buffer.c:144)
 by 0x48AD987: av_buffer_unref (buffer.c:139)
 by 0x5D06D7A: avcodec_close (avcodec.c:486)
 by 0x628DD7D: avcodec_free_context (options.c:175)
 by 0x10A863: main (main.cpp:115)
 Address 0x17812700 is 0 bytes inside a block of size 24 free'd
 at 0x484488F: free (vg_replace_malloc.c:985)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:127)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:144)
 by 0x48AD98F: av_buffer_unref (buffer.c:139)
 by 0x48BE098: hwframe_ctx_free (hwcontext.c:240)
 by 0x48AD9A6: UnknownInlinedFun (buffer.c:133)
 by 0x48AD9A6: UnknownInlinedFun (buffer.c:144)
 by 0x48AD9A6: av_buffer_unref (buffer.c:139)
 by 0x5D06D0A: UnknownInlinedFun (decode.c:1261)
 by 0x5D06D0A: avcodec_close (avcodec.c:465)
 by 0x628DD7D: avcodec_free_context (options.c:175)
 by 0x10A863: main (main.cpp:115)
 Block was alloc'd at
 at 0x4849366: posix_memalign (vg_replace_malloc.c:2099)
 by 0x48D9BD5: av_malloc (mem.c:105)
 by 0x48D9DAD: av_mallocz (mem.c:256)
 by 0x48AD8DD: UnknownInlinedFun (buffer.c:44)
 by 0x48AD8DD: av_buffer_create (buffer.c:64)
 by 0x48BDDEB: av_hwdevice_ctx_alloc (hwcontext.c:179)
 by 0x48BDF29: av_hwdevice_ctx_create (hwcontext.c:622)
 by 0x10A482: main (main.cpp:43)

Invalid free() / delete / delete[] / realloc()
 at 0x484488F: free (vg_replace_malloc.c:985)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:127)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:144)
 by 0x48AD98F: av_buffer_unref (buffer.c:139)
 by 0x5D06D7A: avcodec_close (avcodec.c:486)
 by 0x628DD7D: avcodec_free_context (options.c:175)
 by 0x10A863: main (main.cpp:115)
 Address 0x17812700 is 0 bytes inside a block of size 24 free'd
 at 0x484488F: free (vg_replace_malloc.c:985)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:127)
 by 0x48AD98F: UnknownInlinedFun (buffer.c:144)
 by 0x48AD98F: av_buffer_unref (buffer.c:139)
 by 0x48BE098: hwframe_ctx_free (hwcontext.c:240)
 by 0x48AD9A6: UnknownInlinedFun (buffer.c:133)
 by 0x48AD9A6: UnknownInlinedFun (buffer.c:144)
 by 0x48AD9A6: av_buffer_unref (buffer.c:139)
 by 0x5D06D0A: UnknownInlinedFun (decode.c:1261)
 by 0x5D06D0A: avcodec_close (avcodec.c:465)
 by 0x628DD7D: avcodec_free_context (options.c:175)
 by 0x10A863: main (main.cpp:115)
 Block was alloc'd at
 at 0x4849366: posix_memalign (vg_replace_malloc.c:2099)
 by 0x48D9BD5: av_malloc (mem.c:105)
 by 0x48D9DAD: av_mallocz (mem.c:256)
 by 0x48AD8DD: UnknownInlinedFun (buffer.c:44)
 by 0x48AD8DD: av_buffer_create (buffer.c:64)
 by 0x48BDDEB: av_hwdevice_ctx_alloc (hwcontext.c:179)
 by 0x48BDF29: av_hwdevice_ctx_create (hwcontext.c:622)
 by 0x10A482: main (main.cpp:43)


that is also duplicated (due to the calls to
avcodec_free_context()andav_buffer_unref()).

The question is : How can I fix this ?


The (more or less) minimal (not) working example reads


#include <string>

extern "C" {
 #include <libavutil></libavutil>opt.h>
 #include <libavcodec></libavcodec>avcodec.h>
 #include <libavformat></libavformat>avformat.h>
 #include <libavutil></libavutil>hwcontext.h>
 #include <libavutil></libavutil>pixdesc.h>
 #include <libavutil></libavutil>hwcontext_cuda.h>
}

//(former) libx264 encoding based on https://github.com/FFmpeg/FFmpeg/blob/master/doc/examples/muxing.c
//update to h264_nvenc with a lot of help from https://stackoverflow.com/questions/49862610/opengl-to-ffmpeg-encode
//and some additional info of https://github.com/FFmpeg/FFmpeg/blob/master/doc/examples/vaapi_encode.c

int main() {
 const int _SrcImageWidth=640;
 const int _SrcImageHeight=480;
 
 const AVOutputFormat *_oFmt = nullptr;
 AVFormatContext *_oFmtCtx = nullptr;
 
 const AVCodec *_pCodec = nullptr;
 AVCodecContext *_pCodecCtx = nullptr;
 
 AVFrame* _frame;
 AVPacket* _packet;
 AVStream* _stream;
 
 AVBufferRef *_hwDeviceRefCtx = nullptr;
 const CUcontext* _cudaCtx;
 
 const std::string _OutFileName = "output.mkv";
 
 //constructor part
 int ret;

 //output format context 
 avformat_alloc_output_context2( &_oFmtCtx, nullptr, nullptr, _OutFileName.c_str() );
 _oFmt = _oFmtCtx->oformat;

 //hardware format context
 ret = av_hwdevice_ctx_create( &_hwDeviceRefCtx, AV_HWDEVICE_TYPE_CUDA, "NVIDIA GeForce RTX 4070", nullptr, 0 );

 //hardware frame context for device buffer allocation
 AVBufferRef* hwFrameRefCtx = av_hwframe_ctx_alloc( _hwDeviceRefCtx );
 AVHWFramesContext* hwFrameCtx = (AVHWFramesContext*) (hwFrameRefCtx->data);
 hwFrameCtx->width = _SrcImageWidth;
 hwFrameCtx->height = _SrcImageHeight;
 hwFrameCtx->sw_format = AV_PIX_FMT_0BGR32;
 hwFrameCtx->format = AV_PIX_FMT_CUDA;
 hwFrameCtx->device_ref = _hwDeviceRefCtx;
 hwFrameCtx->device_ctx = (AVHWDeviceContext*) _hwDeviceRefCtx->data;

 ret = av_hwframe_ctx_init( hwFrameRefCtx );

 //get cuda context
 const AVHWDeviceContext* hwDeviceCtx = (AVHWDeviceContext*)(_hwDeviceRefCtx->data);
 const AVCUDADeviceContext* cudaDeviceCtx = (AVCUDADeviceContext*)(hwDeviceCtx->hwctx);
 _cudaCtx = &(cudaDeviceCtx->cuda_ctx);

 //codec context
 _pCodec = avcodec_find_encoder_by_name( "h264_nvenc" );

 _packet = av_packet_alloc();

 _stream = avformat_new_stream( _oFmtCtx, nullptr );
 _stream->id = _oFmtCtx->nb_streams - 1;
 _pCodecCtx = avcodec_alloc_context3( _pCodec );

 _pCodecCtx->qmin = 18;
 _pCodecCtx->qmax = 20;
 _pCodecCtx->width = _SrcImageWidth;
 _pCodecCtx->height = _SrcImageHeight;
 _pCodecCtx->framerate = (AVRational) {25,1};
 _pCodecCtx->time_base = (AVRational) {1,25};
 _stream->time_base = _pCodecCtx->time_base;
 _pCodecCtx->gop_size = 12; //I-Frame every at most 12 frames
 _pCodecCtx->max_b_frames = 2;
 _pCodecCtx->pix_fmt = AV_PIX_FMT_CUDA; //required to use renderbuffer as src
 _pCodecCtx->codec_type = AVMEDIA_TYPE_VIDEO;
 _pCodecCtx->sw_pix_fmt = AV_PIX_FMT_0BGR32; 
 _pCodecCtx->hw_device_ctx = _hwDeviceRefCtx;
 _pCodecCtx->hw_frames_ctx = av_buffer_ref( hwFrameRefCtx );
 av_opt_set(_pCodecCtx->priv_data, "preset", "p7", 0);
 av_opt_set(_pCodecCtx->priv_data, "rc", "vbr", 0);
 if( _oFmtCtx->oformat->flags & AVFMT_GLOBALHEADER ) {
 _pCodecCtx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
 }

 ret = avcodec_open2( _pCodecCtx, _pCodec, nullptr );
 avcodec_parameters_from_context( _stream->codecpar, _pCodecCtx );

 if (!(_oFmtCtx->oformat->flags & AVFMT_NOFILE)) {
 ret = avio_open(&_oFmtCtx->pb, _OutFileName.c_str(), AVIO_FLAG_WRITE);
 }
 ret = avformat_write_header( _oFmtCtx, nullptr );

 //use hardware frame from above
 _frame = av_frame_alloc();
 ret = av_hwframe_get_buffer( _pCodecCtx->hw_frames_ctx, _frame, 0 );
 _frame->pts = 1;

 av_buffer_unref( &hwFrameRefCtx );

 //destructor part
 av_frame_free( &_frame );
 av_packet_free( &_packet );

 av_write_trailer( _oFmtCtx );
 avio_closep( &_oFmtCtx->pb );

 avformat_free_context( _oFmtCtx );

 avcodec_free_context( &_pCodecCtx );
 av_buffer_unref( &_hwDeviceRefCtx );

 return 0;
}
</string>

and compiles with (linux user)


g++ -lavutil -lavformat -lavcodec -lz -lavutil -lswscale -lswresample -lm -ggdb3 -I/opt/cuda/include main.cpp


Thanks in advance !


Syndiquer tout le site
