I was looking for a tutorial/book that would teach me how to start to use FFmpeg as a library (a.k.a. libav) and then I found the tutorial. Unfortunately it was deprecated, so I decided to write this one.

Most of the code in here will be in c but don’t worry: you can easily understand and apply it to your preferred language. FFmpeg libav has lots of bindings for many languages like python, and even if your language doesn’t have it, you can still support it through the (here’s an example with Lua).

We’ll start with a quick lesson about what is video, audio, codec and container and then we’ll go to a crash course on how to use FFmpeg command line and finally we’ll write code, feel free to skip directly to the section Learn FFmpeg libav the Hard Way.

Some people used to say that the Internet video streaming is the future of the traditional TV, in any case, the FFmpeg is something that is worth studying.

Intro

If you have a sequence series of images and change them at a given frequency (let’s say 24 images per second), you will create an . In summary this is the very basic idea behind a video: a series of pictures / frames running at a given rate.

Zeitgenössische Illustration (1886)

audio - what you listen!

Although a muted video can express a variety of feelings, adding sound to it brings more pleasure to the experience.

Sound is the vibration that propagates as a wave of pressure, through the air or any other transmission medium, such as a gas, liquid or solid.

audio analog to digital

codec - shrinking data

CODEC is an electronic circuit or software that compresses or decompresses digital audio/video. It converts raw (uncompressed) digital audio/video to a compressed format or vice versa.

But if we chose to pack millions of images in a single file and called it a movie, we might end up with a huge file. Let’s do the math:

Suppose we are creating a video with a resolution of 1080 x 1920 (height x width) and that we’ll spend 3 bytes per pixel (the minimal point at a screen) to encode the color (or 24 bit color, what gives us 16,777,216 different colors) and this video runs at 24 frames per second and it is 30 minutes long.

This video would require approximately 250.28GB of storage or 1.11Gbps of bandwidth! That’s why we need to use a .

container - a comfy place for audio and video

A container or wrapper format is a metafile format whose specification describes how different elements of data and metadata coexist in a computer file.

A single file that contains all the streams (mostly the audio and video) and it also provides synchronization and general metadata, such as title, resolution and etc.

Usually we can infer the format of a file by looking at its extension: for instance a video.webm is probably a video using the container webm.

FFmpeg - command line

A complete, cross-platform solution to record, convert and stream audio and video.

To work with multimedia we can use the AMAZING tool/library called . Chances are you already know/use it directly or indirectly (do you use Chrome?).

It has a command line program called ffmpeg, a very simple yet powerful binary. For instance, you can convert from mp4 to the container avi just by typing the follow command:

  1. $ ffmpeg -i input.mp4 output.avi

We just made a remuxing here, which is converting from one container to another one. Technically FFmpeg could also be doing a transcoding but we’ll talk about that later.

FFmpeg command line tool 101

FFmpeg does have a documentation that does a great job of explaining how it works.

To make things short, the FFmpeg command line program expects the following argument format to perform its actions ffmpeg {1} {2} -i {3} {4} {5}, where:

  1. global options
  2. input file options
  3. input url
  4. output file options
  5. output url

The parts 2, 3, 4 and 5 can be as many as you need. It’s easier to understand this argument format in action:

  1. # WARNING: this file is around 300MB
  2. $ wget -O bunny_1080p_60fps.mp4 http://distribution.bbb3d.renderfarming.net/video/mp4/bbb_sunflower_1080p_60fps_normal.mp4
  3.  
  4. $ ffmpeg \
  5. -y \ # global options
  6. -c:a libfdk_aac -c:v libx264 \ # input options
  7. -i bunny_1080p_60fps.mp4 \ # input url
  8. -c:v libvpx-vp9 -c:a libvorbis \ # output options
  9. bunny_1080p_60fps_vp9.webm # output url

This command takes an input file mp4 containing two streams (an audio encoded with aac CODEC and a video encoded using h264 CODEC) and convert it to webm, changing its audio and video CODECs too.

We could simplify the command above but then be aware that FFmpeg will adopt or guess the default values for you. For instance when you just type ffmpeg -i input.avi output.mp4 what audio/video CODEC does it use to produce the output.mp4?

Werner Robitza wrote a must read/execute .

Common video operations

While working with audio/video we usually do a set of tasks with the media.

What? the act of converting one of the streams (audio or video) from one CODEC to another one.

Why? sometimes some devices (TVs, smartphones, console and etc) doesn’t support X but Y and newer CODECs provide better compression rate.

How? converting an H264 (AVC) video to an H265 (HEVC).

  1. $ ffmpeg \
  2. -i bunny_1080p_60fps.mp4 \
  3. -c:v libx265 \
  4. bunny_1080p_60fps_h265.mp4

Transmuxing

What? the act of converting from one format (container) to another one.

Why? sometimes some devices (TVs, smartphones, console and etc) doesn’t support X but Y and sometimes newer containers provide modern required features.

How? converting a mp4 to a webm.

  1. $ ffmpeg \
  2. -i bunny_1080p_60fps.mp4 \
  3. -c copy \ # just saying to ffmpeg to skip encoding
  4. bunny_1080p_60fps.webm

Transrating

What? the act of changing the bit rate, or producing other renditions.

Why? people will try to watch your video in a 2G (edge) connection using a less powerful smartphone or in a fiber Internet connection on their 4K TVs therefore you should offer more than one rendition of the same video with different bit rate.

How? producing a rendition with bit rate between 3856K and 2000K.

  1. $ ffmpeg \
  2. -i bunny_1080p_60fps.mp4 \
  3. -minrate 964K -maxrate 3856K -bufsize 2000K \
  4. bunny_1080p_60fps_transrating_964_3856.mp4

Usually we’ll be using transrating with transsizing. Werner Robitza wrote another must read/execute series of posts about FFmpeg rate control.

Transsizing

What? the act of converting from one resolution to another one. As said before transsizing is often used with transrating.

How? converting a 1080p to a 480p resolution.

  1. $ ffmpeg \
  2. -i bunny_1080p_60fps.mp4 \
  3. -vf scale=480:-1 \
  4. bunny_1080p_60fps_transsizing_480.mp4

Bonus Round: Adaptive Streaming

adaptive streaming

What? the act of producing many resolutions (bit rates) and split the media into chunks and serve them via http.

Why? to provide a flexible media that can be watched on a low end smartphone or on a 4K TV, it’s also easy to scale and deploy but it can add latency.

How? creating an adaptive WebM using DASH.

  1. # video streams
  2. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:v libvpx-vp9 -s 160x90 -b:v 250k -keyint_min 150 -g 150 -an -f webm -dash 1 video_160x90_250k.webm
  3.  
  4. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:v libvpx-vp9 -s 320x180 -b:v 500k -keyint_min 150 -g 150 -an -f webm -dash 1 video_320x180_500k.webm
  5.  
  6. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:v libvpx-vp9 -s 640x360 -b:v 750k -keyint_min 150 -g 150 -an -f webm -dash 1 video_640x360_750k.webm
  7.  
  8. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:v libvpx-vp9 -s 640x360 -b:v 1000k -keyint_min 150 -g 150 -an -f webm -dash 1 video_640x360_1000k.webm
  9.  
  10. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:v libvpx-vp9 -s 1280x720 -b:v 1500k -keyint_min 150 -g 150 -an -f webm -dash 1 video_1280x720_1500k.webm
  11.  
  12. # audio streams
  13. $ ffmpeg -i bunny_1080p_60fps.mp4 -c:a libvorbis -b:a 128k -vn -f webm -dash 1 audio_128k.webm
  14.  
  15. # the DASH manifest
  16. $ ffmpeg \
  17.  -f webm_dash_manifest -i video_160x90_250k.webm \
  18.  -f webm_dash_manifest -i video_320x180_500k.webm \
  19.  -f webm_dash_manifest -i video_640x360_750k.webm \
  20.  -f webm_dash_manifest -i video_640x360_1000k.webm \
  21.  -f webm_dash_manifest -i video_1280x720_500k.webm \
  22.  -f webm_dash_manifest -i audio_128k.webm \
  23.  -c copy -map 0 -map 1 -map 2 -map 3 -map 4 -map 5 \
  24.  -f webm_dash_manifest \
  25.  -adaptation_sets "id=0,streams=0,1,2,3,4 id=1,streams=5" \
  26.  manifest.mpd

PS: I stole this example from the

There are many and many other usages for FFmpeg. I use it in conjunction with iMovie to produce/edit some videos for YouTube and you can certainly use it professionally.

Learn FFmpeg libav the Hard Way

Since the FFmpeg is so useful as a command line tool to do essential tasks over the media files, how can we use it in our programs?

FFmpeg is that can be integrated into our own programs. Usually, when you install FFmpeg, it installs automatically all these libraries. I’ll be referring to the set of these libraries as FFmpeg libav.

This title is a homage to Zed Shaw’s series Learn X the Hard Way, particularly his book Learn C the Hard Way.

Chapter 0 - The infamous hello world

This hello world actually won’t show the message "hello world" in the terminal 👅 Instead we’re going to print out information about the video, things like its format (container), duration, resolution, audio channels and, in the end, we’ll decode some frames and save them as image files.

But before we start to code, let’s learn how FFmpeg libav architecture works and how its components communicate with others.

Here’s a diagram of the process of decoding a video:

You’ll first need to load your media file into a component called (the video container is also known as format). It actually doesn’t fully load the whole file: it often only reads the header.

Once we loaded the minimal header of our container, we can access its streams (think of them as a rudimentary audio and video data). Each stream will be available in a component called AVStream.

Stream is a fancy name for a continuous flow of data.

Suppose our video has two streams: an audio encoded with and a video encoded with H264 (AVC) CODEC. From each stream we can extract pieces (slices) of data called packets that will be loaded into components named .

The data inside the packets are still coded (compressed) and in order to decode the packets, we need to pass them to a specific AVCodec.

The AVCodec will decode them into and finally, this component gives us the uncompressed frame. Noticed that the same terminology/process is used either by audio and video stream.

Since some people were facing issues while compiling or running the examples we’re going to use as our development/runner environment, we’ll also use the big buck bunny video so if you don’t have it locally just run the command make fetch_small_bunny_video.

TLDR; show me the code and execution.

  1. $ make run_hello

We’ll skip some details, but don’t worry: the .

We’re going to allocate memory to the component AVFormatContext that will hold information about the format (container).

  1. AVFormatContext *pFormatContext = avformat_alloc_context();

Now we’re going to open the file and read its header and fill the AVFormatContext with minimal information about the format (notice that usually the codecs are not opened). The function used to do this is . It expects an AVFormatContext, a filename and two optional arguments: the AVInputFormat (if you pass NULL, FFmpeg will guess the format) and the (which are the options to the demuxer).

  1. avformat_open_input(&pFormatContext, filename, NULL, NULL);

We can print the format name and the media duration:

  1. printf("Format %s, duration %lld us", pFormatContext->iformat->long_name, pFormatContext->duration);

To access the streams, we need to read data from the media. The function avformat_find_stream_info does that. Now, the pFormatContext->nb_streams will hold the amount of streams and the pFormatContext->streams[i] will give us the i stream (an ).

  1. avformat_find_stream_info(pFormatContext,  NULL);

Now we’ll loop through all the streams.

  1. for (int i = 0; i < pFormatContext->nb_streams; i++)
  2. {
  3.   //
  4. }

For each stream, we’re going to keep the AVCodecParameters, which describes the properties of a codec used by the stream i.

  1. AVCodecParameters *pLocalCodecParameters = pFormatContext->streams[i]->codecpar;

With the codec properties we can look up the proper CODEC querying the function and find the registered decoder for the codec id and return an , the component that knows how to enCOde and DECode the stream.

Now we can print information about the codecs.

  1. // specific for video and audio
  2. if (pLocalCodecParameters->codec_type == AVMEDIA_TYPE_VIDEO) {
  3.   printf("Video Codec: resolution %d x %d", pLocalCodecParameters->width, pLocalCodecParameters->height);
  4. } else if (pLocalCodecParameters->codec_type == AVMEDIA_TYPE_AUDIO) {
  5.   printf("Audio Codec: %d channels, sample rate %d", pLocalCodecParameters->channels, pLocalCodecParameters->sample_rate);
  6. }
  7. // general
  8. printf("\tCodec %s ID %d bit_rate %lld", pLocalCodec->long_name, pLocalCodec->id, pCodecParameters->bit_rate);

With the codec, we can allocate memory for the , which will hold the context for our decode/encode process, but then we need to fill this codec context with CODEC parameters; we do that with avcodec_parameters_to_context.

Once we filled the codec context, we need to open the codec. We call the function and then we can use it.

  1. AVCodecContext *pCodecContext = avcodec_alloc_context3(pCodec);
  2. avcodec_parameters_to_context(pCodecContext, pCodecParameters);
  3. avcodec_open2(pCodecContext, pCodec, NULL);

Now we’re going to read the packets from the stream and decode them into frames but first, we need to allocate memory for both components, the AVPacket and .

  1. AVPacket *pPacket = av_packet_alloc();
  2. AVFrame *pFrame = av_frame_alloc();

Let’s feed our packets from the streams with the function av_read_frame while it has packets.

  1. while (av_read_frame(pFormatContext, pPacket) >= 0) {
  2.   //...
  3. }

Let’s send the raw data packet (compressed frame) to the decoder, through the codec context, using the function .

  1. avcodec_send_packet(pCodecContext, pPacket);

And let’s receive the raw data frame (uncompressed frame) from the decoder, through the same codec context, using the function avcodec_receive_frame.

  1. avcodec_receive_frame(pCodecContext, pFrame);

We can print the frame number, the , DTS, frame type and etc.

  1. printf(
  2.     "Frame %c (%d) pts %d dts %d key_frame %d [coded_picture_number %d, display_picture_number %d]",
  3.     av_get_picture_type_char(pFrame->pict_type),
  4.     pCodecContext->frame_number,
  5.     pFrame->pts,
  6.     pFrame->pkt_dts,
  7.     pFrame->key_frame,
  8.     pFrame->coded_picture_number,
  9.     pFrame->display_picture_number
  10. );

Finally we can save our decoded frame into a . The process is very simple, we’ll use the pFrame->data where the index is related to the planes Y, Cb and Cr, we just picked 0 (Y) to save our gray image.

  1. save_gray_frame(pFrame->data[0], pFrame->linesize[0], pFrame->width, pFrame->height, frame_filename);
  2.  
  3. static void save_gray_frame(unsigned char *buf, int wrap, int xsize, int ysize, char *filename)
  4. {
  5.     FILE *f;
  6.     f = fopen(filename,"w");
  7.     // writing the minimal required header for a pgm file format
  8.     // portable graymap format -> https://en.wikipedia.org/wiki/Netpbm_format#PGM_example
  9.     fprintf(f, "P5\n%d %d\n%d\n", xsize, ysize, 255);
  10.  
  11.     // writing line by line
  12.     for (i = 0; i < ysize; i++)
  13.         fwrite(buf + i * wrap, 1, xsize, f);
  14.     fclose(f);
  15. }

And voilà! Now we have a gray scale image with 2MB:

Chapter 1 - syncing audio and video

Be the player - a young JS developer writing a new MSE video player.

Before we move to let’s talk about timing, or how a video player knows the right time to play a frame.

In the last example, we saved some frames that can be seen here:

When we’re designing a video player we need to play each frame at a given pace, otherwise it would be hard to pleasantly see the video either because it’s playing so fast or so slow.

It’s easier to understand when we look at some examples, let’s simulate some scenarios.

For a fps=60/1 and timebase=1/60000 each PTS will increase timescale / fps = 1000 therefore the PTS real time for each frame could be (supposing it started at 0):

  • frame=0, PTS = 0, PTS_TIME = 0
  • frame=1, PTS = 1000, PTS_TIME = PTS * timebase = 0.016
  • frame=2, PTS = 2000, PTS_TIME = PTS * timebase = 0.033

For almost the same scenario but with a timebase equal to 1/60.

  • frame=0, PTS = 0, PTS_TIME = 0
  • frame=1, PTS = 1, PTS_TIME = PTS * timebase = 0.016
  • frame=2, PTS = 2, PTS_TIME = PTS * timebase = 0.033
  • frame=3, PTS = 3, PTS_TIME = PTS * timebase = 0.050

For a fps=25/1 and timebase=1/75 each PTS will increase timescale / fps = 3 and the PTS time could be:

  • frame=0, PTS = 0, PTS_TIME = 0
  • frame=1, PTS = 3, PTS_TIME = PTS * timebase = 0.04
  • frame=2, PTS = 6, PTS_TIME = PTS * timebase = 0.08
  • frame=3, PTS = 9, PTS_TIME = PTS * timebase = 0.12
  • …
  • frame=24, PTS = 72, PTS_TIME = PTS * timebase = 0.96
  • …
  • frame=4064, PTS = 12192, PTS_TIME = PTS * timebase = 162.56

Now with the pts_time we can find a way to render this synched with audio pts_time or with a system clock. The FFmpeg libav provides these info through its API:

Just out of curiosity, the frames we saved were sent in a DTS order (frames: 1,6,4,2,3,5) but played at a PTS order (frames: 1,2,3,4,5). Also, notice how cheap are B-Frames in comparison to P or I-Frames.

  1. LOG: AVStream->r_frame_rate 60/1
  2. LOG: AVStream->time_base 1/60000
  3. ...
  4. LOG: Frame 1 (type=I, size=153797 bytes) pts 6000 key_frame 1 [DTS 0]
  5. LOG: Frame 2 (type=B, size=8117 bytes) pts 7000 key_frame 0 [DTS 3]
  6. LOG: Frame 3 (type=B, size=8226 bytes) pts 8000 key_frame 0 [DTS 4]
  7. LOG: Frame 4 (type=B, size=17699 bytes) pts 9000 key_frame 0 [DTS 2]
  8. LOG: Frame 5 (type=B, size=6253 bytes) pts 10000 key_frame 0 [DTS 5]
  9. LOG: Frame 6 (type=P, size=34992 bytes) pts 11000 key_frame 0 [DTS 1]

Chapter 2 - remuxing

Remuxing is the act of changing from one format (container) to another, for instance, we can change a MPEG-4 video to a one without much pain using FFmpeg:

  1. ffmpeg input.mp4 -c copy output.ts

It’ll demux the mp4 but it won’t decode or encode it (-c copy) and in the end, it’ll mux it into a mpegts file. If you don’t provide the format -f the ffmpeg will try to guess it based on the file’s extension.

The general usage of FFmpeg or the libav follows a pattern/architecture or workflow:

  • protocol layer - it accepts an input (a file for instance but it could be a rtmp or HTTP input as well)
  • - it demuxes its content, revealing mostly metadata and its streams
  • codec layer - it decodes its compressed streams data optional
  • - it can also apply some to the raw frames (like resizing)optional
  • and then it does the reverse path
  • codec layer - it encodes (or re-encodes or even transcodes) the raw framesoptional
  • - it muxes (or remuxes) the raw streams (the compressed data)
  • protocol layer - and finally the muxed data is sent to an output (another file or maybe a network remote server)

Now let’s code an example using libav to provide the same effect as in ffmpeg input.mp4 -c copy output.ts.

We’re going to read from an input (input_format_context) and change it to another output (output_format_context).

  1. AVFormatContext *input_format_context = NULL;
  2. AVFormatContext *output_format_context = NULL;

We start doing the usually allocate memory and open the input format. For this specific case, we’re going to open an input file and allocate memory for an output file.

  1. if ((ret = avformat_open_input(&input_format_context, in_filename, NULL, NULL)) < 0) {
  2.   fprintf(stderr, "Could not open input file '%s'", in_filename);
  3.   goto end;
  4. }
  5. if ((ret = avformat_find_stream_info(input_format_context, NULL)) < 0) {
  6.   fprintf(stderr, "Failed to retrieve input stream information");
  7.   goto end;
  8. }
  9.  
  10. avformat_alloc_output_context2(&output_format_context, NULL, NULL, out_filename);
  11. if (!output_format_context) {
  12.   fprintf(stderr, "Could not create output context\n");
  13.   ret = AVERROR_UNKNOWN;
  14.   goto end;
  15. }

We’re going to remux only the video, audio and subtitle types of streams so we’re holding what streams we’ll be using into an array of indexes.

  1. number_of_streams = input_format_context->nb_streams;
  2. streams_list = av_mallocz_array(number_of_streams, sizeof(*streams_list));

Just after we allocated the required memory, we’re going to loop throughout all the streams and for each one we need to create new out stream into our output format context, using the avformat_new_stream function. Notice that we’re marking all the streams that aren’t video, audio or subtitle so we can skip them after.

  1. for (i = 0; i < input_format_context->nb_streams; i++) {
  2.   AVStream *out_stream;
  3.   AVStream *in_stream = input_format_context->streams[i];
  4.   AVCodecParameters *in_codecpar = in_stream->codecpar;
  5.   if (in_codecpar->codec_type != AVMEDIA_TYPE_AUDIO &&
  6.       in_codecpar->codec_type != AVMEDIA_TYPE_VIDEO &&
  7.       in_codecpar->codec_type != AVMEDIA_TYPE_SUBTITLE) {
  8.     streams_list[i] = -1;
  9.     continue;
  10.   }
  11.   streams_list[i] = stream_index++;
  12.   out_stream = avformat_new_stream(output_format_context, NULL);
  13.   if (!out_stream) {
  14.     fprintf(stderr, "Failed allocating output stream\n");
  15.     ret = AVERROR_UNKNOWN;
  16.     goto end;
  17.   }
  18.   ret = avcodec_parameters_copy(out_stream->codecpar, in_codecpar);
  19.   if (ret < 0) {
  20.     fprintf(stderr, "Failed to copy codec parameters\n");
  21.     goto end;
  22.   }
  23. }

Now we can create the output file.

After that, we can copy the streams, packet by packet, from our input to our output streams. We’ll loop while it has packets (av_read_frame), for each packet we need to re-calculate the PTS and DTS to finally write it (av_interleaved_write_frame) to our output format context.

  1. while (1) {
  2.   AVStream *in_stream, *out_stream;
  3.   ret = av_read_frame(input_format_context, &packet);
  4.   if (ret < 0)
  5.     break;
  6.   in_stream  = input_format_context->streams[packet.stream_index];
  7.   if (packet.stream_index >= number_of_streams || streams_list[packet.stream_index] < 0) {
  8.     av_packet_unref(&packet);
  9.     continue;
  10.   }
  11.   packet.stream_index = streams_list[packet.stream_index];
  12.   out_stream = output_format_context->streams[packet.stream_index];
  13.   /* copy packet */
  14.   packet.pts = av_rescale_q_rnd(packet.pts, in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
  15.   packet.dts = av_rescale_q_rnd(packet.dts, in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
  16.   packet.duration = av_rescale_q(packet.duration, in_stream->time_base, out_stream->time_base);
  17.   // https://ffmpeg.org/doxygen/trunk/structAVPacket.html#ab5793d8195cf4789dfb3913b7a693903
  18.   packet.pos = -1;
  19.  
  20.   //https://ffmpeg.org/doxygen/trunk/group__lavf__encoding.html#ga37352ed2c63493c38219d935e71db6c1
  21.   ret = av_interleaved_write_frame(output_format_context, &packet);
  22.   if (ret < 0) {
  23.     fprintf(stderr, "Error muxing packet\n");
  24.     break;
  25.   }
  26.   av_packet_unref(&packet);
  27. }

To finalize we need to write the stream trailer to an output media file with function.

  1. av_write_trailer(output_format_context);

Now we’re ready to test it and the first test will be a format (video container) conversion from a MP4 to a MPEG-TS video file. We’re basically making the command line ffmpeg input.mp4 -c copy output.ts with libav.

  1. make run_remuxing_ts

It’s working!!! don’t you trust me?! you shouldn’t, we can check it with ffprobe:

  1. ffprobe -i remuxed_small_bunny_1080p_60fps.ts
  2.  
  3. Input #0, mpegts, from 'remuxed_small_bunny_1080p_60fps.ts':
  4.   Duration: 00:00:10.03, start: 0.000000, bitrate: 2751 kb/s
  5.   Program 1
  6.     Metadata:
  7.       service_name    : Service01
  8.       service_provider: FFmpeg
  9.     Stream #0:1[0x101]: Audio: ac3 ([129][0][0][0] / 0x0081), 48000 Hz, 5.1(side), fltp, 320 kb/s

To sum up what we did here in a graph, we can revisit our initial idea about how libav works but showing that we skipped the codec part.

Before we end this chapter I’d like to show an important part of the remuxing process, you can pass options to the muxer. Let’s say we want to delivery MPEG-DASH format for that matter we need to use (sometimes referred as fmp4) instead of MPEG-TS or plain MPEG-4.

With the command line we can do that easily.

  1. ffmpeg -i non_fragmented.mp4 -movflags frag_keyframe+empty_moov+default_base_moof fragmented.mp4

Almost equally easy as the command line is the libav version of it, we just need to pass the options when write the output header, just before the packets copy.

  1. AVDictionary* opts = NULL;
  2. av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov+default_base_moof", 0);
  3. ret = avformat_write_header(output_format_context, &opts);

We now can generate this fragmented mp4 file:

  1. make run_remuxing_fragmented_mp4

But to make sure that I’m not lying to you. You can use the amazing site/tool or the site http://mp4parser.com/ to see the differences, first load up the “common” mp4.

As you can see it has a single mdat atom/box, this is place where the video and audio frames are. Now load the fragmented mp4 to see which how it spreads the mdat boxes.

fragmented mp4 boxes

Chapter 3 - transcoding

TLDR; show me the code and execution.

  1. $ make run_transcoding

We’ll skip some details, but don’t worry: the .

In this chapter, we’re going to create a minimalist transcoder, written in C, that can convert videos coded in H264 to H265 using FFmpeg/libav library specifically libavcodec, libavformat, and libavutil.

Just a quick recap: The AVFormatContext is the abstraction for the format of the media file, aka container (ex: MKV, MP4, Webm, TS). The represents each type of data for a given format (ex: audio, video, subtitle, metadata). The AVPacket is a slice of compressed data obtained from the AVStream that can be decoded by an (ex: av1, h264, vp9, hevc) generating a raw data called AVFrame.

Let’s start with the simple transmuxing operation and then we can build upon this code, the first step is to load the input file.

  1. // Allocate an AVFormatContext
  2. avfc = avformat_alloc_context();
  3. // Open an input stream and read the header.
  4. avformat_open_input(avfc, in_filename, NULL, NULL);
  5. // Read packets of a media file to get stream information.
  6. avformat_find_stream_info(avfc, NULL);

Now we’re going to set up the decoder, the AVFormatContext will give us access to all the AVStream components and for each one of them, we can get their AVCodec and create the particular AVCodecContext and finally we can open the given codec so we can proceed to the decoding process.

The holds data about media configuration such as bit rate, frame rate, sample rate, channels, height, and many others.

  1. for (int i = 0; i < avfc->nb_streams; i++)
  2. {
  3.   AVStream *avs = avfc->streams[i];
  4.   AVCodec *avc = avcodec_find_decoder(avs->codecpar->codec_id);
  5.   AVCodecContext *avcc = avcodec_alloc_context3(*avc);
  6.   avcodec_parameters_to_context(*avcc, avs->codecpar);
  7.   avcodec_open2(*avcc, *avc, NULL);
  8. }

We need to prepare the output media file for transmuxing as well, we first allocate memory for the output AVFormatContext. We create each stream in the output format. In order to pack the stream properly, we copy the codec parameters from the decoder.

We set the flag AV_CODEC_FLAG_GLOBAL_HEADER which tells the encoder that it can use the global headers and finally we open the output file for write and persist the headers.

  1. avformat_alloc_output_context2(&encoder_avfc, NULL, NULL, out_filename);
  2.  
  3. AVStream *avs = avformat_new_stream(encoder_avfc, NULL);
  4. avcodec_parameters_copy(avs->codecpar, decoder_avs->codecpar);
  5.  
  6. if (encoder_avfc->oformat->flags & AVFMT_GLOBALHEADER)
  7.   encoder_avfc->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
  8.  
  9. avio_open(&encoder_avfc->pb, encoder->filename, AVIO_FLAG_WRITE);
  10. avformat_write_header(encoder->avfc, &muxer_opts);

We’re getting the AVPacket‘s from the decoder, adjusting the timestamps, and write the packet properly to the output file. Even though the function av_interleaved_write_frame says “write frame” we are storing the packet. We finish the transmuxing process by writing the stream trailer to the file.

  1. AVFrame *input_frame = av_frame_alloc();
  2. AVPacket *input_packet = av_packet_alloc();
  3.  
  4. while (av_read_frame(decoder_avfc, input_packet) >= 0)
  5. {
  6.   av_packet_rescale_ts(input_packet, decoder_video_avs->time_base, encoder_video_avs->time_base);
  7.   av_interleaved_write_frame(*avfc, input_packet) < 0));
  8. }
  9.  
  10. av_write_trailer(encoder_avfc);

The previous section showed a simple transmuxer program, now we’re going to add the capability to encode files, specifically we’re going to enable it to transcode videos from h264 to h265.

After we prepared the decoder but before we arrange the output media file we’re going to set up the encoder.

  1. AVRational input_framerate = av_guess_frame_rate(decoder_avfc, decoder_video_avs, NULL);
  2. AVStream *video_avs = avformat_new_stream(encoder_avfc, NULL);
  3.  
  4. char *codec_name = "libx265";
  5. char *codec_priv_key = "x265-params";
  6. // we're going to use internal options for the x265
  7. // it disables the scene change detection and fix then
  8. // GOP on 60 frames.
  9. char *codec_priv_value = "keyint=60:min-keyint=60:scenecut=0";
  10.  
  11. AVCodec *video_avc = avcodec_find_encoder_by_name(codec_name);
  12. AVCodecContext *video_avcc = avcodec_alloc_context3(video_avc);
  13. // encoder codec params
  14. av_opt_set(sc->video_avcc->priv_data, codec_priv_key, codec_priv_value, 0);
  15. video_avcc->height = decoder_ctx->height;
  16. video_avcc->width = decoder_ctx->width;
  17. video_avcc->pix_fmt = video_avc->pix_fmts[0];
  18. // control rate
  19. video_avcc->bit_rate = 2 * 1000 * 1000;
  20. video_avcc->rc_buffer_size = 4 * 1000 * 1000;
  21. video_avcc->rc_max_rate = 2 * 1000 * 1000;
  22. video_avcc->rc_min_rate = 2.5 * 1000 * 1000;
  23. // time base
  24. video_avcc->time_base = av_inv_q(input_framerate);
  25. video_avs->time_base = sc->video_avcc->time_base;
  26.  
  27. avcodec_open2(sc->video_avcc, sc->video_avc, NULL);
  28. avcodec_parameters_from_context(sc->video_avs->codecpar, sc->video_avcc);

We need to expand our decoding loop for the video stream transcoding:

  1. AVFrame *input_frame = av_frame_alloc();
  2. AVPacket *input_packet = av_packet_alloc();
  3.  
  4. while (av_read_frame(decoder_avfc, input_packet) >= 0)
  5. {
  6.   int response = avcodec_send_packet(decoder_video_avcc, input_packet);
  7.   while (response >= 0) {
  8.     response = avcodec_receive_frame(decoder_video_avcc, input_frame);
  9.     if (response == AVERROR(EAGAIN) || response == AVERROR_EOF) {
  10.       break;
  11.     } else if (response < 0) {
  12.       return response;
  13.     }
  14.     if (response >= 0) {
  15.       encode(encoder_avfc, decoder_video_avs, encoder_video_avs, decoder_video_avcc, input_packet->stream_index);
  16.     }
  17.     av_frame_unref(input_frame);
  18.   }
  19.   av_packet_unref(input_packet);
  20. }
  21. av_write_trailer(encoder_avfc);
  22.  
  23. // used function
  24. int encode(AVFormatContext *avfc, AVStream *dec_video_avs, AVStream *enc_video_avs, AVCodecContext video_avcc int index) {
  25.   AVPacket *output_packet = av_packet_alloc();
  26.   int response = avcodec_send_frame(video_avcc, input_frame);
  27.  
  28.   while (response >= 0) {
  29.     response = avcodec_receive_packet(video_avcc, output_packet);
  30.     if (response == AVERROR(EAGAIN) || response == AVERROR_EOF) {
  31.       break;
  32.     } else if (response < 0) {
  33.       return -1;
  34.     }
  35.  
  36.     output_packet->stream_index = index;
  37.     output_packet->duration = enc_video_avs->time_base.den / enc_video_avs->time_base.num / dec_video_avs->avg_frame_rate.num * dec_video_avs->avg_frame_rate.den;
  38.  
  39.     av_packet_rescale_ts(output_packet, dec_video_avs->time_base, enc_video_avs->time_base);
  40.     response = av_interleaved_write_frame(avfc, output_packet);
  41.   }
  42.   av_packet_unref(output_packet);
  43.   av_packet_free(&output_packet);
  44.   return 0;

We converted the media stream from h264 to h265, as expected the h265 version of the media file is smaller than the h264 however the is capable of:

Now, to be honest, this was harder than I thought it’d be and I had to dig into the and test it a lot and I think I’m missing something because I had to enforce force-cfr for the h264 to work and I’m still seeing some warning messages like warning messages (forced frame type (5) at 80 was changed to frame type (3)).