[FFmpeg-devel] HLS Segmenter and the "hls_time" option
Steven Liu
lingjiujianke at gmail.com
Fri Aug 26 12:06:29 EEST 2016
2016-08-26 16:58 GMT+08:00 Ibrahim Tachijian <barhom at netsat.se>:
> Thanks this actually does what I expected it to do.
> For me this option will help a lot, and we would really be interested in
> this eventually reaching git.
>
> Thanks to Steven Liu's patch we will be able to use this to start live
> streams faster. Great job!
>
> Below follows what command I ran and the output m3u8 playlist.
>
> /root/ffmpeg_patched -analyzeduration 1000000 -i udp://MCAST_ADDR:3301 -map
> 0:0 -map 0:1 -c:v libx264 -preset superfast -g 25 -b:v 900k -maxrate 900k
> -bufsize 2000k -filter:v yadif -c:a libfdk_aac -b:a 64k -hls_time 5
> *-hls_init_time
> 1* -hls_list_size 10 -hls_allow_cache 0 -hls_flags delete_segments -f hls
> /tmp/playlist.m3u8
>
> The output M3U8 of patched ffmpeg:
>
> #EXTM3U
> #EXT-X-VERSION:3
> #EXT-X-ALLOW-CACHE:NO
> #EXT-X-TARGETDURATION:5
> #EXT-X-MEDIA-SEQUENCE:3
> #EXTINF:1.000000,
> playlist3.ts
> #EXTINF:1.000000,
> playlist4.ts
> #EXTINF:1.000000,
> playlist5.ts
> #EXTINF:1.000000,
> playlist6.ts
> #EXTINF:1.000000,
> playlist7.ts
> #EXTINF:1.000000,
> playlist8.ts
> #EXTINF:1.600000,
> playlist9.ts
> #EXTINF:1.000000,
> playlist10.ts
> #EXTINF:4.000000,
> playlist11.ts
> #EXTINF:5.000000,
> playlist12.ts
>
>
> On Fri, Aug 26, 2016 at 8:37 AM Steven Liu <lingjiujianke at gmail.com>
> wrote:
>
> > 2016-08-26 14:00 GMT+08:00 Steven Liu <lingjiujianke at gmail.com>:
> >
> > >
> > >
> > > 2016-08-26 10:34 GMT+08:00 Ibrahim Tachijian <barhom at netsat.se>:
> > >
> > >> In my use case scenario I only need it for the very first couple of
> > >> segments.
> > >> After 5 segments it is not a problem anymore to have 5 second segments
> > >> only.
> > >>
> > >>
> > >>
> > >> On Fri, Aug 26, 2016 at 4:25 AM Steven Liu <lingjiujianke at gmail.com>
> > >> wrote:
> > >>
> > >> > 2016-08-26 10:10 GMT+08:00 Ibrahim Tachijian <barhom at netsat.se>:
> > >> >
> > >> > > yes that is correct Steven.
> > >> > >
> > >> > > On Fri, Aug 26, 2016 at 3:41 AM Steven Liu <
> lingjiujianke at gmail.com
> > >
> > >> > > wrote:
> > >> > >
> > >> > > > 2016-08-26 8:17 GMT+08:00 Ibrahim Tachijian <barhom at netsat.se>:
> > >> > > >
> > >> > > > > Steven, I am not sure you understood me correctly or perhaps I
> > did
> > >> > not
> > >> > > > > explain myself optimally.
> > >> > > > >
> > >> > > > > We still want to split by keyframe in a normal fashion. But,
> for
> > >> > > example,
> > >> > > > > would like the first 5 segments to have an "hls_time" of 1s
> and
> > >> the
> > >> > > rest
> > >> > > > of
> > >> > > > > the segments after the first 5 to have an "hls_time" of 5s.
> > >> > > > >
> > >> > > > > An made up option would be (hls_time_initial Seconds,Number)
> > >> > > > >
> > >> > > > > - "-hls_time_initial 1,5 -hls_time 5"
> > >> > > > >
> > >> > > > > The output playlist would contain segments (split at
> keyframes)
> > 5
> > >> > > > segments
> > >> > > > > of length 1s and then any segment after the initial 5 segment
> > >> would
> > >> > be
> > >> > > > 5s.
> > >> > > > >
> > >> > > > > Is it clear what I am trying to explain?
> > >> > > > >
> > >> > > > > What do you think? Do you know how this can be achieved?
> > >> > > > >
> > >> > > > > Thanks,
> > >> > > > >
> > >> > > > > On Fri, Aug 26, 2016 at 2:06 AM Steven Liu <
> > >> lingjiujianke at gmail.com>
> > >> > > > > wrote:
> > >> > > > >
> > >> > > > > > 2016-08-26 7:39 GMT+08:00 Ibrahim Tachijian <
> barhom at netsat.se
> > >:
> > >> > > > > >
> > >> > > > > > > Hello,
> > >> > > > > > >
> > >> > > > > > > I've been thinking about an option for "hls_time" that is
> > not
> > >> > > > currently
> > >> > > > > > > supported by FFMpeg and I would like feedback to if some
> of
> > >> you
> > >> > may
> > >> > > > > think
> > >> > > > > > > this is useful or utterly unnecessary.
> > >> > > > > > >
> > >> > > > > > > I find scenarios where we sometimes want to create an HLS
> > >> output
> > >> > > and
> > >> > > > > > would
> > >> > > > > > > like the *first couple of segments* to be shorter than the
> > >> *rest
> > >> > of
> > >> > > > the
> > >> > > > > > > segments.*
> > >> > > > > > >
> > >> > > > > > > For example starting off with 1s segments up to N segments
> > >> then
> > >> > > > switch
> > >> > > > > to
> > >> > > > > > > 5s segments.
> > >> > > > > > >
> > >> > > > > > > The reasoning is simply to have the playlist.m3u8 and
> first
> > >> > segment
> > >> > > > > > > available *asap.*
> > >> > > > > > >
> > >> > > > > > > What do you think? Do you know how this can be achieved?
> > >> > > > > > >
> > >> > > > > > > Thanks,
> > >> > > > > > >
> > >> > > > > >
> > >> > > > > > ffmpeg split segment use hlsenc by keyframe, if you want to
> > >> split
> > >> > > > segment
> > >> > > > > > *ASAP*,
> > >> > > > > > you can try set your streaming *GOP(Group Of Picture)*
> > shorter,
> > >> > > > > > for example: gop_size = 60, c->time_base =
> > (AVRational){1,60};
> > >> > > > > > in command line : -r:v 60 -g 60
> > >> > > > > >
> > >> > > > > > if can accept split segment at non-keyframe, you can use
> > >> hls_time
> > >> > and
> > >> > > > > > hls_flags split_by_time.
> > >> > > > > > _______________________________________________
> > >> > > > >
> > >> > > > you mean:
> > >> > > >
> > >> > > > segment-0.ts,segment-1.ts,segment-2.ts,segment-3.ts,
> segment-4.ts
> > >> these
> > >> > > 5
> > >> > > > segments duration is 1s
> > >> > > > and
> > >> > > > segment-5.ts,segment-6.ts,segment-7.ts ...... these segment
> > >> duration is
> > >> > > 5
> > >> > > > or set by hls_time
> > >> > > > Is that right i understand?
> > >> > >
> > >> > >
> > >> > OK , Thanks.
> > >> > That a good idea, but i have a question:
> > >> > when ffmpeg update the segment list after 10 segments, there have a
> > new
> > >> > user to play the hls playlist, the m3u8 segments every ts segments'
> > >> > duration is the 5s.
> > >> > So do you want exec a new ffmpeg when every one play the hls
> playlist
> > ?
> > >>
> > >>
> > > Dose this patch can fix your request?
> > >
> > >
> > > update patch
> >
> >
> > Test process:
> > localhost:ffmpeg liuqi$ ./ffmpeg -v verbose -re -i
> > ~/Movies/objectC/facebook.mp4 -c copy -f hls -hls_time 3 -hls_list_size 3
> > -hls_flags split_by_time aaaa_output.m3u8
> > ffmpeg version N-81462-gbcd1153 Copyright (c) 2000-2016 the FFmpeg
> > developers
> > built with Apple LLVM version 7.3.0 (clang-703.0.31)
> > configuration: --enable-fontconfig --enable-gpl --enable-libass
> > --enable-libbluray --enable-libfreetype --enable-libmp3lame
> > --enable-libspeex --enable-libx264 --enable-libx265 --enable-version3
> > libavutil 55. 29.100 / 55. 29.100
> > libavcodec 57. 54.100 / 57. 54.100
> > libavformat 57. 48.100 / 57. 48.100
> > libavdevice 57. 0.102 / 57. 0.102
> > libavfilter 6. 55.100 / 6. 55.100
> > libswscale 4. 1.100 / 4. 1.100
> > libswresample 2. 1.100 / 2. 1.100
> > libpostproc 54. 0.100 / 54. 0.100
> > [h264 @ 0x7f94f1813c00] Reinit context to 1280x720, pix_fmt: yuv420p
> > Input #0, mov,mp4,m4a,3gp,3g2,mj2, from
> > '/Users/liuqi/Movies/objectC/facebook.mp4':
> > Metadata:
> > major_brand : isom
> > minor_version : 1
> > compatible_brands: isomavc1
> > creation_time : 2015-02-02 18:19:19
> > Duration: 00:45:02.06, start: 0.000000, bitrate: 2708 kb/s
> > Stream #0:0(und): Video: h264 (High), 1 reference frame (avc1 /
> > 0x31637661), yuv420p, 1280x714 (1280x720) [SAR 1:1 DAR 640:357], 2576
> kb/s,
> > 25 fps, 25 tbr, 25k tbn, 50 tbc (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:19
> > handler_name : GPAC ISO Video Handler
> > Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz,
> > stereo, fltp, 127 kb/s (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:23
> > handler_name : GPAC ISO Audio Handler
> > [hls @ 0x7f94f3004800] Using AVStream.codec to pass codec parameters to
> > muxers is deprecated, use AVStream.codecpar instead.
> > Last message repeated 1 times
> > [mpegts @ 0x7f94f180b800] muxrate VBR, pcr every 5 pkts, sdt every
> > 2147483647, pat/pmt every 2147483647 pkts
> > Output #0, hls, to 'aaaa_output.m3u8':
> > Metadata:
> > major_brand : isom
> > minor_version : 1
> > compatible_brands: isomavc1
> > encoder : Lavf57.48.100
> > Stream #0:0(und): Video: h264 (High), 1 reference frame (avc1 /
> > 0x31637661), yuv420p, 1280x714 (0x0) [SAR 1:1 DAR 640:357], q=2-31, 2576
> > kb/s, 25 fps, 25 tbr, 90k tbn, 50 tbc (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:19
> > handler_name : GPAC ISO Video Handler
> > Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz,
> > stereo, 127 kb/s (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:23
> > handler_name : GPAC ISO Audio Handler
> > Stream mapping:
> > Stream #0:0 -> #0:0 (copy)
> > Stream #0:1 -> #0:1 (copy)
> > Press [q] to stop, [?] for help
> > Automatically inserted bitstream filter 'h264_mp4toannexb'; args=''
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:00:03.04 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:00:06.12 bitrate=N/A
> > speed=1.01x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:00:08.64 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:10:11.73 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:20:14.74 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:30:17.76 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:40:20.84 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:50:23.89 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7f94f3004800] EXT-X-MEDIA-SEQUENCE:6
> > frame= 601 fps= 25 q=-1.0 Lsize=N/A time=00:00:24.29 bitrate=N/A speed=
> > 1x
> > video:7099kB audio:370kB subtitle:0kB other streams:0kB global
> headers:0kB
> > muxing overhead: unknown
> > Input file #0 (/Users/liuqi/Movies/objectC/facebook.mp4):
> > Input stream #0:0 (video): 601 packets read (7269480 bytes);
> > Input stream #0:1 (audio): 1139 packets read (379370 bytes);
> > Total: 1740 packets (7648850 bytes) demuxed
> > Output file #0 (aaaa_output.m3u8):
> > Output stream #0:0 (video): 601 packets muxed (7269480 bytes);
> > Output stream #0:1 (audio): 1139 packets muxed (379370 bytes);
> > Total: 1740 packets (7648850 bytes) muxed
> > localhost:ffmpeg liuqi$
> > localhost:ffmpeg liuqi$ ./ffmpeg -v verbose -re -i
> > ~/Movies/objectC/facebook.mp4 -c copy -f hls -hls_init_time 1 -hls_time 5
> > -hls_flags split_by_time output_test.m3u8
> > ffmpeg version N-81462-gbcd1153 Copyright (c) 2000-2016 the FFmpeg
> > developers
> > built with Apple LLVM version 7.3.0 (clang-703.0.31)
> > configuration: --enable-fontconfig --enable-gpl --enable-libass
> > --enable-libbluray --enable-libfreetype --enable-libmp3lame
> > --enable-libspeex --enable-libx264 --enable-libx265 --enable-version3
> > libavutil 55. 29.100 / 55. 29.100
> > libavcodec 57. 54.100 / 57. 54.100
> > libavformat 57. 48.100 / 57. 48.100
> > libavdevice 57. 0.102 / 57. 0.102
> > libavfilter 6. 55.100 / 6. 55.100
> > libswscale 4. 1.100 / 4. 1.100
> > libswresample 2. 1.100 / 2. 1.100
> > libpostproc 54. 0.100 / 54. 0.100
> > [h264 @ 0x7fedf4017600] Reinit context to 1280x720, pix_fmt: yuv420p
> > Input #0, mov,mp4,m4a,3gp,3g2,mj2, from
> > '/Users/liuqi/Movies/objectC/facebook.mp4':
> > Metadata:
> > major_brand : isom
> > minor_version : 1
> > compatible_brands: isomavc1
> > creation_time : 2015-02-02 18:19:19
> > Duration: 00:45:02.06, start: 0.000000, bitrate: 2708 kb/s
> > Stream #0:0(und): Video: h264 (High), 1 reference frame (avc1 /
> > 0x31637661), yuv420p, 1280x714 (1280x720) [SAR 1:1 DAR 640:357], 2576
> kb/s,
> > 25 fps, 25 tbr, 25k tbn, 50 tbc (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:19
> > handler_name : GPAC ISO Video Handler
> > Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz,
> > stereo, fltp, 127 kb/s (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:23
> > handler_name : GPAC ISO Audio Handler
> > [hls @ 0x7fedf4000600] Using AVStream.codec to pass codec parameters to
> > muxers is deprecated, use AVStream.codecpar instead.
> > Last message repeated 1 times
> > [mpegts @ 0x7fedf3008a00] muxrate VBR, pcr every 5 pkts, sdt every
> > 2147483647, pat/pmt every 2147483647 pkts
> > Output #0, hls, to 'output_test.m3u8':
> > Metadata:
> > major_brand : isom
> > minor_version : 1
> > compatible_brands: isomavc1
> > encoder : Lavf57.48.100
> > Stream #0:0(und): Video: h264 (High), 1 reference frame (avc1 /
> > 0x31637661), yuv420p, 1280x714 (0x0) [SAR 1:1 DAR 640:357], q=2-31, 2576
> > kb/s, 25 fps, 25 tbr, 90k tbn, 50 tbc (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:19
> > handler_name : GPAC ISO Video Handler
> > Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz,
> > stereo, 127 kb/s (default)
> > Metadata:
> > creation_time : 2015-02-02 18:19:23
> > handler_name : GPAC ISO Audio Handler
> > Stream mapping:
> > Stream #0:0 -> #0:0 (copy)
> > Stream #0:1 -> #0:1 (copy)
> > Press [q] to stop, [?] for help
> > Automatically inserted bitstream filter 'h264_mp4toannexb'; args=''
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:00:00.52 bitrate=N/A
> > speed=1.02x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:00:01.56 bitrate=N/A
> > speed=1.03x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:00:03.08 bitrate=N/A
> > speed=1.01x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:00:03.56 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:00:05.12 bitrate=N/A
> > speed=1.01x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:10:06.12 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:20:09.70 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:30:14.80 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:40:19.88 bitrate=N/A speed=
> > 1x
> > [hls @ 0x7fedf4000600] EXT-X-MEDIA-SEQUENCE:50:20.90 bitrate=N/A speed=
> > 1x
> > frame= 529 fps= 25 q=-1.0 Lsize=N/A time=00:00:21.35 bitrate=N/A speed=
> > 1x
> > video:6506kB audio:324kB subtitle:0kB other streams:0kB global
> headers:0kB
> > muxing overhead: unknown
> > Input file #0 (/Users/liuqi/Movies/objectC/facebook.mp4):
> > Input stream #0:0 (video): 529 packets read (6661710 bytes);
> > Input stream #0:1 (audio): 1001 packets read (332091 bytes);
> > Total: 1530 packets (6993801 bytes) demuxed
> > Output file #0 (output_test.m3u8):
> > Output stream #0:0 (video): 529 packets muxed (6661710 bytes);
> > Output stream #0:1 (audio): 1001 packets muxed (332091 bytes);
> > Total: 1530 packets (6993801 bytes) muxed
> > localhost:ffmpeg liuqi$
> >
> >
> > And recover segments duration time by hls_time after init hls window.
> > This is reuqested by Ibrahim Tachijian
> >
> > Signed-off-by: LiuQi <liuqi at gosun.com>
> > ---
> > libavformat/hlsenc.c | 14 +++++++++++++-
> > 1 file changed, 13 insertions(+), 1 deletion(-)
> >
> > diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
> > index e65f002..f5ceb60 100644
> > --- a/libavformat/hlsenc.c
> > +++ b/libavformat/hlsenc.c
> > @@ -85,6 +85,7 @@ typedef struct HLSContext {
> > AVFormatContext *vtt_avf;
> >
> > float time; // Set by a private option.
> > + float init_time; // Set by a private option.
> > int max_nb_segments; // Set by a private option.
> > int wrap; // Set by a private option.
> > uint32_t flags; // enum HLSFlags
> > @@ -706,7 +707,7 @@ static int hls_write_header(AVFormatContext *s)
> > int vtt_basename_size;
> >
> > hls->sequence = hls->start_sequence;
> > - hls->recording_time = hls->time * AV_TIME_BASE;
> > + hls->recording_time = (hls->init_time ? hls->init_time : hls->time)
> *
> > AV_TIME_BASE;
> > hls->start_pts = AV_NOPTS_VALUE;
> >
> > if (hls->format_options_str) {
> > @@ -860,9 +861,19 @@ static int hls_write_packet(AVFormatContext *s,
> > AVPacket *pkt)
> > AVStream *st = s->streams[pkt->stream_index];
> > int64_t end_pts = hls->recording_time * hls->number;
> > int is_ref_pkt = 1;
> > + int init_list_dur = 0;
> > + int after_init_list_dur = 0;
> > int ret, can_split = 1;
> > int stream_index = 0;
> >
> > + if (hls->sequence - hls->nb_entries > hls->start_sequence &&
> > hls->init_time > 0) {
> > + /* reset end_pts, hls->recording_time at end of the init hls
> list
> > */
> > + init_list_dur = hls->init_time * hls->nb_entries * AV_TIME_BASE;
> > + after_init_list_dur = (hls->sequence - hls->nb_entries ) *
> > hls->time * AV_TIME_BASE;
> > + hls->recording_time = hls->time * AV_TIME_BASE;
> > + end_pts = init_list_dur + after_init_list_dur ;
> > + }
> > +
> > if( st->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE ) {
> > oc = hls->vtt_avf;
> > stream_index = 0;
> > @@ -972,6 +983,7 @@ static int hls_write_trailer(struct AVFormatContext
> *s)
> > static const AVOption options[] = {
> > {"start_number", "set first number in the sequence",
> > OFFSET(start_sequence),AV_OPT_TYPE_INT64, {.i64 = 0}, 0,
> INT64_MAX,
> > E},
> > {"hls_time", "set segment length in seconds",
> > OFFSET(time), AV_OPT_TYPE_FLOAT, {.dbl = 2}, 0, FLT_MAX, E},
> > + {"hls_init_time", "set segment length in seconds at init list",
> > OFFSET(init_time), AV_OPT_TYPE_FLOAT, {.dbl = 0}, 0, FLT_MAX,
> > E},
> > {"hls_list_size", "set maximum number of playlist entries",
> > OFFSET(max_nb_segments), AV_OPT_TYPE_INT, {.i64 = 5}, 0,
> > INT_MAX, E},
> > {"hls_ts_options","set hls mpegts list of options for the container
> > format used for hls", OFFSET(format_options_str), AV_OPT_TYPE_STRING,
> {.str
> > = NULL}, 0, 0, E},
> > {"hls_vtt_options","set hls vtt list of options for the container
> > format used for hls", OFFSET(vtt_format_options_str),
> AV_OPT_TYPE_STRING,
> > {.str = NULL}, 0, 0, E},
> > --
> > 2.7.4 (Apple Git-66)
>
>
Ok, This function look good to me to. i will sent this patch to maillist in
a new mail.
More information about the ffmpeg-devel
mailing list