[FFmpeg-devel] [PATCH] libavformat/matroskaenc: Add WebM DASH support
Michael Niedermayer
michaelni at gmx.at
Wed May 14 00:50:48 CEST 2014
On Tue, May 13, 2014 at 03:30:19PM -0700, Vignesh Venkatasubramanian wrote:
> On Tue, May 13, 2014 at 10:07 AM, Vignesh Venkatasubramanian
> <vigneshv at google.com> wrote:
> > On Tue, May 13, 2014 at 6:08 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
> >> On Mon, May 12, 2014 at 11:25:08PM -0700, Vignesh Venkatasubramanian wrote:
> >>> On Thu, May 8, 2014 at 10:11 AM, Vignesh Venkatasubramanian
> >>> <vigneshv at google.com> wrote:
> >>> > On Wed, May 7, 2014 at 3:37 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> >>> >> On Wed, May 07, 2014 at 02:32:26PM -0700, Vignesh Venkatasubramanian wrote:
> >>> >>> On Wed, May 7, 2014 at 1:40 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> >>> >>> > On Wed, May 07, 2014 at 09:46:40AM -0700, Vignesh Venkatasubramanian wrote:
> >>> >>> >> WebM DASH specification [1] requires the Clusters and Cues to be output in a
> >>> >>> >> specific way. Adding a flag to matroskaenc that will enable support for
> >>> >>> >> creating WebM/Mkv files conforming to the WebM DASH specification.
> >>> >>> >>
> >>> >>> >> [1] http://wiki.webmproject.org/adaptive-streaming/webm-dash-specification
> >>> >>> >>
> >>> >>> >> Signed-off-by: Vignesh Venkatasubramanian <vigneshv at google.com>
> >>> >>> >> ---
> >>> >>> >> libavformat/matroskaenc.c | 68 ++++++++++++++++++++++++++++++++++-------------
> >>> >>> >> 1 file changed, 50 insertions(+), 18 deletions(-)
> >>> >>> >>
> >>> >>> >> diff --git a/libavformat/matroskaenc.c b/libavformat/matroskaenc.c
> >>> >>> >> index e5e6116..3cf6e73 100644
> >>> >>> >> --- a/libavformat/matroskaenc.c
> >>> >>> >> +++ b/libavformat/matroskaenc.c
> >>> >>> >> @@ -70,6 +70,7 @@ typedef struct mkv_seekhead {
> >>> >>> >>
> >>> >>> >> typedef struct {
> >>> >>> >> uint64_t pts;
> >>> >>> >> + int stream_idx;
> >>> >>> >> int tracknum;
> >>> >>> >> int64_t cluster_pos; ///< file offset of the cluster containing the block
> >>> >>> >> int64_t relative_pos; ///< relative offset from the position of the cluster containing the block
> >>> >>> >> @@ -114,6 +115,8 @@ typedef struct MatroskaMuxContext {
> >>> >>> >> int cluster_size_limit;
> >>> >>> >> int64_t cues_pos;
> >>> >>> >> int64_t cluster_time_limit;
> >>> >>> >> + int is_dash;
> >>> >>> >> + int dash_track_number;
> >>> >>> >>
> >>> >>> >> uint32_t chapter_id_offset;
> >>> >>> >> int wrote_chapters;
> >>> >>> >> @@ -399,8 +402,8 @@ static mkv_cues * mkv_start_cues(int64_t segment_offset)
> >>> >>> >> return cues;
> >>> >>> >> }
> >>> >>> >>
> >>> >>> >> -static int mkv_add_cuepoint(mkv_cues *cues, int stream, int64_t ts, int64_t cluster_pos, int64_t relative_pos,
> >>> >>> >> - int64_t duration)
> >>> >>> >> +static int mkv_add_cuepoint(mkv_cues *cues, int stream, int tracknum, int64_t ts,
> >>> >>> >> + int64_t cluster_pos, int64_t relative_pos, int64_t duration)
> >>> >>> >> {
> >>> >>> >> mkv_cuepoint *entries = cues->entries;
> >>> >>> >>
> >>> >>> >> @@ -413,7 +416,8 @@ static int mkv_add_cuepoint(mkv_cues *cues, int stream, int64_t ts, int64_t clus
> >>> >>> >> cues->entries = entries;
> >>> >>> >>
> >>> >>> >> cues->entries[cues->num_entries].pts = ts;
> >>> >>> >> - cues->entries[cues->num_entries].tracknum = stream + 1;
> >>> >>> >> + cues->entries[cues->num_entries].stream_idx = stream;
> >>> >>> >> + cues->entries[cues->num_entries].tracknum = tracknum;
> >>> >>> >> cues->entries[cues->num_entries].cluster_pos = cluster_pos - cues->segment_offset;
> >>> >>> >> cues->entries[cues->num_entries].relative_pos = relative_pos;
> >>> >>> >> cues->entries[cues->num_entries++].duration = duration;
> >>> >>> >> @@ -443,7 +447,7 @@ static int64_t mkv_write_cues(AVIOContext *pb, mkv_cues *cues, mkv_track *tracks
> >>> >>> >> for (j = 0; j < num_tracks; j++)
> >>> >>> >> tracks[j].has_cue = 0;
> >>> >>> >> for (j = 0; j < cues->num_entries - i && entry[j].pts == pts; j++) {
> >>> >>> >> - int tracknum = entry[j].tracknum - 1;
> >>> >>> >> + int tracknum = entry[j].stream_idx;
> >>> >>> >> av_assert0(tracknum>=0 && tracknum<num_tracks);
> >>> >>> >> if (tracks[tracknum].has_cue)
> >>> >>> >> continue;
> >>> >>> >> @@ -646,8 +650,10 @@ static int mkv_write_tracks(AVFormatContext *s)
> >>> >>> >> get_aac_sample_rates(s, codec, &sample_rate, &output_sample_rate);
> >>> >>> >>
> >>> >>> >> track = start_ebml_master(pb, MATROSKA_ID_TRACKENTRY, 0);
> >>> >>> >> - put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER , i + 1);
> >>> >>> >> - put_ebml_uint (pb, MATROSKA_ID_TRACKUID , i + 1);
> >>> >>> >> + put_ebml_uint (pb, MATROSKA_ID_TRACKNUMBER,
> >>> >>> >> + mkv->is_dash ? mkv->dash_track_number : i + 1);
> >>> >>> >> + put_ebml_uint (pb, MATROSKA_ID_TRACKUID,
> >>> >>> >> + mkv->is_dash ? mkv->dash_track_number : i + 1);
> >>> >>> >> put_ebml_uint (pb, MATROSKA_ID_TRACKFLAGLACING , 0); // no lacing (yet)
> >>> >>> >>
> >>> >>> >> if ((tag = av_dict_get(st->metadata, "title", NULL, 0)))
> >>> >>> >> @@ -1425,7 +1431,8 @@ static void mkv_write_block(AVFormatContext *s, AVIOContext *pb,
> >>> >>> >>
> >>> >>> >> put_ebml_id(pb, blockid);
> >>> >>> >> put_ebml_num(pb, size+4, 0);
> >>> >>> >> - avio_w8(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
> >>> >>> >> + // this assumes stream_index is less than 126
> >>> >>> >> + avio_w8(pb, 0x80 | (mkv->is_dash ? mkv->dash_track_number : (pkt->stream_index + 1)));
> >>> >>> >> avio_wb16(pb, ts - mkv->cluster_pts);
> >>> >>> >> avio_w8(pb, flags);
> >>> >>> >> avio_write(pb, data + offset, size);
> >>> >>> >> @@ -1539,7 +1546,7 @@ static void mkv_flush_dynbuf(AVFormatContext *s)
> >>> >>> >> mkv->dyn_bc = NULL;
> >>> >>> >> }
> >>> >>> >>
> >>> >>> >> -static int mkv_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
> >>> >>> >> +static int mkv_write_packet_internal(AVFormatContext *s, AVPacket *pkt, int add_cue)
> >>> >>> >> {
> >>> >>> >> MatroskaMuxContext *mkv = s->priv_data;
> >>> >>> >> AVIOContext *pb = s->pb;
> >>> >>> >> @@ -1596,8 +1603,13 @@ static int mkv_write_packet_internal(AVFormatContext *s, AVPacket *pkt)
> >>> >>> >> end_ebml_master(pb, blockgroup);
> >>> >>> >> }
> >>> >>> >>
> >>> >>> >> - if ((codec->codec_type == AVMEDIA_TYPE_VIDEO && keyframe) || codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
> >>> >>> >> - ret = mkv_add_cuepoint(mkv->cues, pkt->stream_index, ts, mkv->cluster_pos, relative_packet_pos,
> >>> >>> >> + if ((codec->codec_type == AVMEDIA_TYPE_VIDEO && keyframe) ||
> >>> >>> >> + codec->codec_type == AVMEDIA_TYPE_SUBTITLE ||
> >>> >>> >> + add_cue) {
> >>> >>> >> + ret = mkv_add_cuepoint(mkv->cues,
> >>> >>> >> + pkt->stream_index,
> >>> >>> >> + mkv->is_dash ? mkv->dash_track_number : pkt->stream_index + 1,
> >>> >>> >> + ts, mkv->cluster_pos, relative_packet_pos,
> >>> >>> >> codec->codec_type == AVMEDIA_TYPE_SUBTITLE ? duration : -1);
> >>> >>> >> if (ret < 0) return ret;
> >>> >>> >> }
> >>> >>> >> @@ -1615,6 +1627,7 @@ static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
> >>> >>> >> int64_t cluster_time;
> >>> >>> >> AVIOContext *pb;
> >>> >>> >> int ret;
> >>> >>> >> + int start_new_cluster;
> >>> >>> >>
> >>> >>> >> if (mkv->tracks[pkt->stream_index].write_dts)
> >>> >>> >> cluster_time = pkt->dts - mkv->cluster_pts;
> >>> >>> >> @@ -1632,11 +1645,26 @@ static int mkv_write_packet(AVFormatContext *s, AVPacket *pkt)
> >>> >>> >> cluster_size = avio_tell(pb);
> >>> >>> >> }
> >>> >>> >>
> >>> >>> >> - if (mkv->cluster_pos != -1 &&
> >>> >>> >> - (cluster_size > mkv->cluster_size_limit ||
> >>> >>> >> - cluster_time > mkv->cluster_time_limit ||
> >>> >>> >> - (codec_type == AVMEDIA_TYPE_VIDEO && keyframe &&
> >>> >>> >> - cluster_size > 4 * 1024))) {
> >>> >>> >> + if (mkv->is_dash && codec_type == AVMEDIA_TYPE_VIDEO) {
> >>> >>> >> + // WebM DASH specification states that the first block of every cluster
> >>> >>> >> + // has to be a key frame. So for DASH video, we only create a cluster
> >>> >>> >> + // on seeing key frames.
> >>> >>> >
> >>> >>> >> + start_new_cluster = keyframe;
> >>> >>> >
> >>> >>> > is it intended to have 1 frame per cluster for intra only videos ?
> >>> >>>
> >>> >>> yes. but this only done when -dash parameter is explicitly set. so
> >>> >>> it's very unlikely that someone would want intra-only video for dash.
> >>> >>
> >>> >> Someone might already have a intra only video and want to transmit
> >>> >> it without reencoding or
> >>> >> a encoder might choose to encode multiple frames in a row to be
> >>> >> encoded as intra frames (for example in a very high motion scene)
> >>> >>
> >>> >>
> >>> >
> >>> > that looks like a valid use case. i will rather limit it by
> >>> > cluster_time_limit and force a keyframe with the encoder whenever a
> >>> > cluster is to be created. will update the patch soon.
> >>> >
> >>>
> >>> on second thought, using cluster_time_limit as a sole factor will
> >>> produce incorrect DASH files in case of remuxing without encoding (as
> >>> there might not be a key frame when the cluster_time_limit is
> >>> reached).
> >>>
> >>> so there are two options here. one is to use cluster_time_limit as the
> >>> sole factor (and force keyframes with the encode whenever the cluster
> >>> time limit is reached). this will create incorrect files in case of
> >>> remux. the second option is to use the existing patch. this will
> >>> always produce correct DASH files (although it will be inefficient in
> >>> the case of intra-only files).
> >>>
> >>> thoughts?
> >>
> >> I guess we should get it working & compliant first before trying to
> >> make it behave better for less common cases
> >> as this seems not as trivial and obvious to fix as it seemed first
> >>
> >
> > ok fine.
> >
> >> did you had a chance to look at a fate test for this ?
> >
> > working on it. will update the patch when i have it.
> >
>
> is there a way to check for container specific stuff in fate?
> md5/framemd5/crc/framecrc all seem to operate on either individual
> packets (not the container) or incrementally on individual packets.
> since this patch is not going to change the actual contents of the
> packet itself, none of this is useful in this case. i could just use
> md5 on the whole file, but that will be way to fragile (for e.g.,
> muxing application field on the track element is set to lavfxx.yy.zzz
> and it will break when that changes). any other way to test this?
what kind of check would you (ideall) want to have for this ?
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Awnsering whenever a program halts or runs forever is
On a turing machine, in general impossible (turings halting problem).
On any real computer, always possible as a real computer has a finite number
of states N, and will either halt in less than N cycles or never halt.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20140514/952b7f8d/attachment.asc>
More information about the ffmpeg-devel
mailing list