2 * WebM DASH Manifest XML muxer
3 * Copyright (c) 2014 Vignesh Venkatasubramanian
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * WebM DASH Specification:
24 * https://sites.google.com/a/webmproject.org/wiki/adaptive-streaming/webm-dash-specification
25 * ISO DASH Specification:
26 * http://standards.iso.org/ittf/PubliclyAvailableStandards/c065274_ISO_IEC_23009-1_2014.zip
36 #include "libavutil/avstring.h"
37 #include "libavutil/dict.h"
38 #include "libavutil/opt.h"
39 #include "libavutil/time_internal.h"
41 typedef struct AdaptationSet {
47 typedef struct WebMDashMuxContext {
49 char *adaptation_sets;
52 int representation_id;
54 int chunk_start_index;
57 double time_shift_buffer_depth;
58 int minimum_update_period;
61 static const char *get_codec_name(int codec_id)
63 return avcodec_descriptor_get(codec_id)->name;
66 static double get_duration(AVFormatContext *s)
70 for (i = 0; i < s->nb_streams; i++) {
71 AVDictionaryEntry *duration = av_dict_get(s->streams[i]->metadata,
73 if (!duration || atof(duration->value) < 0) continue;
74 if (atof(duration->value) > max) max = atof(duration->value);
79 static int write_header(AVFormatContext *s)
81 WebMDashMuxContext *w = s->priv_data;
82 double min_buffer_time = 1.0;
83 avio_printf(s->pb, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n");
84 avio_printf(s->pb, "<MPD\n");
85 avio_printf(s->pb, " xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n");
86 avio_printf(s->pb, " xmlns=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
87 avio_printf(s->pb, " xsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
88 avio_printf(s->pb, " type=\"%s\"\n", w->is_live ? "dynamic" : "static");
90 avio_printf(s->pb, " mediaPresentationDuration=\"PT%gS\"\n",
93 avio_printf(s->pb, " minBufferTime=\"PT%gS\"\n", min_buffer_time);
94 avio_printf(s->pb, " profiles=\"%s\"%s",
95 w->is_live ? "urn:mpeg:dash:profile:isoff-live:2011" : "urn:webm:dash:profile:webm-on-demand:2012",
96 w->is_live ? "\n" : ">\n");
98 time_t local_time = time(NULL);
100 struct tm *gmt = gmtime_r(&local_time, &gmt_buffer);
102 if (!strftime(gmt_iso, 21, "%Y-%m-%dT%H:%M:%SZ", gmt)) {
103 return AVERROR_UNKNOWN;
105 if (s->flags & AVFMT_FLAG_BITEXACT) {
106 av_strlcpy(gmt_iso, "", 1);
108 avio_printf(s->pb, " availabilityStartTime=\"%s\"\n", gmt_iso);
109 avio_printf(s->pb, " timeShiftBufferDepth=\"PT%gS\"\n", w->time_shift_buffer_depth);
110 avio_printf(s->pb, " minimumUpdatePeriod=\"PT%dS\"", w->minimum_update_period);
111 avio_printf(s->pb, ">\n");
112 if (w->utc_timing_url) {
113 avio_printf(s->pb, "<UTCTiming\n");
114 avio_printf(s->pb, " schemeIdUri=\"urn:mpeg:dash:utc:http-iso:2014\"\n");
115 avio_printf(s->pb, " value=\"%s\"/>\n", w->utc_timing_url);
121 static void write_footer(AVFormatContext *s)
123 avio_printf(s->pb, "</MPD>\n");
126 static int subsegment_alignment(AVFormatContext *s, AdaptationSet *as) {
128 AVDictionaryEntry *gold = av_dict_get(s->streams[as->streams[0]]->metadata,
129 CUE_TIMESTAMPS, NULL, 0);
131 for (i = 1; i < as->nb_streams; i++) {
132 AVDictionaryEntry *ts = av_dict_get(s->streams[as->streams[i]]->metadata,
133 CUE_TIMESTAMPS, NULL, 0);
134 if (!ts || strncmp(gold->value, ts->value, strlen(gold->value))) return 0;
139 static int bitstream_switching(AVFormatContext *s, AdaptationSet *as) {
141 AVDictionaryEntry *gold_track_num = av_dict_get(s->streams[as->streams[0]]->metadata,
142 TRACK_NUMBER, NULL, 0);
143 AVCodecParameters *gold_par = s->streams[as->streams[0]]->codecpar;
144 if (!gold_track_num) return 0;
145 for (i = 1; i < as->nb_streams; i++) {
146 AVDictionaryEntry *track_num = av_dict_get(s->streams[as->streams[i]]->metadata,
147 TRACK_NUMBER, NULL, 0);
148 AVCodecParameters *par = s->streams[as->streams[i]]->codecpar;
150 strncmp(gold_track_num->value, track_num->value, strlen(gold_track_num->value)) ||
151 gold_par->codec_id != par->codec_id ||
152 gold_par->extradata_size != par->extradata_size ||
153 memcmp(gold_par->extradata, par->extradata, par->extradata_size)) {
161 * Writes a Representation within an Adaptation Set. Returns 0 on success and
164 static int write_representation(AVFormatContext *s, AVStream *stream, char *id,
165 int output_width, int output_height,
166 int output_sample_rate) {
167 WebMDashMuxContext *w = s->priv_data;
168 AVDictionaryEntry *irange = av_dict_get(stream->metadata, INITIALIZATION_RANGE, NULL, 0);
169 AVDictionaryEntry *cues_start = av_dict_get(stream->metadata, CUES_START, NULL, 0);
170 AVDictionaryEntry *cues_end = av_dict_get(stream->metadata, CUES_END, NULL, 0);
171 AVDictionaryEntry *filename = av_dict_get(stream->metadata, FILENAME, NULL, 0);
172 AVDictionaryEntry *bandwidth = av_dict_get(stream->metadata, BANDWIDTH, NULL, 0);
173 const char *bandwidth_str;
174 if ((w->is_live && (!filename)) ||
175 (!w->is_live && (!irange || !cues_start || !cues_end || !filename || !bandwidth))) {
176 return AVERROR_INVALIDDATA;
178 avio_printf(s->pb, "<Representation id=\"%s\"", id);
179 // if bandwidth for live was not provided, use a default
180 if (w->is_live && !bandwidth) {
181 bandwidth_str = (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) ? "128000" : "1000000";
183 bandwidth_str = bandwidth->value;
185 avio_printf(s->pb, " bandwidth=\"%s\"", bandwidth_str);
186 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
187 avio_printf(s->pb, " width=\"%d\"", stream->codecpar->width);
188 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
189 avio_printf(s->pb, " height=\"%d\"", stream->codecpar->height);
190 if (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO && output_sample_rate)
191 avio_printf(s->pb, " audioSamplingRate=\"%d\"", stream->codecpar->sample_rate);
193 // For live streams, Codec and Mime Type always go in the Representation tag.
194 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(stream->codecpar->codec_id));
195 avio_printf(s->pb, " mimeType=\"%s/webm\"",
196 stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
197 // For live streams, subsegments always start with key frames. So this
199 avio_printf(s->pb, " startsWithSAP=\"1\"");
200 avio_printf(s->pb, ">");
202 avio_printf(s->pb, ">\n");
203 avio_printf(s->pb, "<BaseURL>%s</BaseURL>\n", filename->value);
204 avio_printf(s->pb, "<SegmentBase\n");
205 avio_printf(s->pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
206 avio_printf(s->pb, "<Initialization\n");
207 avio_printf(s->pb, " range=\"0-%s\" />\n", irange->value);
208 avio_printf(s->pb, "</SegmentBase>\n");
210 avio_printf(s->pb, "</Representation>\n");
215 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
217 static int check_matching_width(AVFormatContext *s, AdaptationSet *as) {
219 if (as->nb_streams < 2) return 1;
220 first_width = s->streams[as->streams[0]]->codecpar->width;
221 for (i = 1; i < as->nb_streams; i++)
222 if (first_width != s->streams[as->streams[i]]->codecpar->width)
228 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
230 static int check_matching_height(AVFormatContext *s, AdaptationSet *as) {
232 if (as->nb_streams < 2) return 1;
233 first_height = s->streams[as->streams[0]]->codecpar->height;
234 for (i = 1; i < as->nb_streams; i++)
235 if (first_height != s->streams[as->streams[i]]->codecpar->height)
241 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
243 static int check_matching_sample_rate(AVFormatContext *s, AdaptationSet *as) {
244 int first_sample_rate, i;
245 if (as->nb_streams < 2) return 1;
246 first_sample_rate = s->streams[as->streams[0]]->codecpar->sample_rate;
247 for (i = 1; i < as->nb_streams; i++)
248 if (first_sample_rate != s->streams[as->streams[i]]->codecpar->sample_rate)
253 static void free_adaptation_sets(AVFormatContext *s) {
254 WebMDashMuxContext *w = s->priv_data;
256 for (i = 0; i < w->nb_as; i++) {
257 av_freep(&w->as[i].streams);
264 * Parses a live header filename and computes the representation id,
265 * initialization pattern and the media pattern. Pass NULL if you don't want to
266 * compute any of those 3. Returns 0 on success and non-zero on failure.
268 * Name of the header file should conform to the following pattern:
269 * <file_description>_<representation_id>.hdr where <file_description> can be
270 * anything. The chunks should be named according to the following pattern:
271 * <file_description>_<representation_id>_<chunk_number>.chk
273 static int parse_filename(char *filename, char **representation_id,
274 char **initialization_pattern, char **media_pattern) {
275 char *underscore_pos = NULL;
276 char *period_pos = NULL;
277 char *temp_pos = NULL;
278 char *filename_str = av_strdup(filename);
282 ret = AVERROR(ENOMEM);
285 temp_pos = av_stristr(filename_str, "_");
287 underscore_pos = temp_pos + 1;
288 temp_pos = av_stristr(temp_pos + 1, "_");
290 if (!underscore_pos) {
291 ret = AVERROR_INVALIDDATA;
294 period_pos = av_stristr(underscore_pos, ".");
296 ret = AVERROR_INVALIDDATA;
299 *(underscore_pos - 1) = 0;
300 if (representation_id) {
301 *representation_id = av_malloc(period_pos - underscore_pos + 1);
302 if (!(*representation_id)) {
303 ret = AVERROR(ENOMEM);
306 av_strlcpy(*representation_id, underscore_pos, period_pos - underscore_pos + 1);
308 if (initialization_pattern) {
309 *initialization_pattern = av_asprintf("%s_$RepresentationID$.hdr",
311 if (!(*initialization_pattern)) {
312 ret = AVERROR(ENOMEM);
317 *media_pattern = av_asprintf("%s_$RepresentationID$_$Number$.chk",
319 if (!(*media_pattern)) {
320 ret = AVERROR(ENOMEM);
326 av_freep(&filename_str);
331 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
333 static int write_adaptation_set(AVFormatContext *s, int as_index)
335 WebMDashMuxContext *w = s->priv_data;
336 AdaptationSet *as = &w->as[as_index];
337 AVCodecParameters *par = s->streams[as->streams[0]]->codecpar;
338 AVDictionaryEntry *lang;
340 static const char boolean[2][6] = { "false", "true" };
341 int subsegmentStartsWithSAP = 1;
343 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
344 // are the same for all contained Representations. otherwise, they will go
345 // on their respective Representation tag. For live streams, they always go
346 // in the Representation tag.
347 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
348 if (par->codec_type == AVMEDIA_TYPE_VIDEO) {
349 width_in_as = !w->is_live && check_matching_width(s, as);
350 height_in_as = !w->is_live && check_matching_height(s, as);
352 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
355 avio_printf(s->pb, "<AdaptationSet id=\"%s\"", as->id);
356 avio_printf(s->pb, " mimeType=\"%s/webm\"",
357 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
358 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(par->codec_id));
360 lang = av_dict_get(s->streams[as->streams[0]]->metadata, "language", NULL, 0);
361 if (lang) avio_printf(s->pb, " lang=\"%s\"", lang->value);
363 if (par->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
364 avio_printf(s->pb, " width=\"%d\"", par->width);
365 if (par->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
366 avio_printf(s->pb, " height=\"%d\"", par->height);
367 if (par->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
368 avio_printf(s->pb, " audioSamplingRate=\"%d\"", par->sample_rate);
370 avio_printf(s->pb, " bitstreamSwitching=\"%s\"",
371 boolean[bitstream_switching(s, as)]);
372 avio_printf(s->pb, " subsegmentAlignment=\"%s\"",
373 boolean[w->is_live || subsegment_alignment(s, as)]);
375 for (i = 0; i < as->nb_streams; i++) {
376 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
377 CLUSTER_KEYFRAME, NULL, 0);
378 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
380 avio_printf(s->pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
381 avio_printf(s->pb, ">\n");
384 AVDictionaryEntry *filename =
385 av_dict_get(s->streams[as->streams[0]]->metadata, FILENAME, NULL, 0);
386 char *initialization_pattern = NULL;
387 char *media_pattern = NULL;
388 int ret = parse_filename(filename->value, NULL, &initialization_pattern,
391 avio_printf(s->pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
392 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
393 avio_printf(s->pb, "<SegmentTemplate");
394 avio_printf(s->pb, " timescale=\"1000\"");
395 avio_printf(s->pb, " duration=\"%d\"", w->chunk_duration);
396 avio_printf(s->pb, " media=\"%s\"", media_pattern);
397 avio_printf(s->pb, " startNumber=\"%d\"", w->chunk_start_index);
398 avio_printf(s->pb, " initialization=\"%s\"", initialization_pattern);
399 avio_printf(s->pb, "/>\n");
400 av_free(initialization_pattern);
401 av_free(media_pattern);
404 for (i = 0; i < as->nb_streams; i++) {
405 char *representation_id = NULL;
408 AVDictionaryEntry *filename =
409 av_dict_get(s->streams[as->streams[i]]->metadata, FILENAME, NULL, 0);
411 return AVERROR(EINVAL);
412 if (ret = parse_filename(filename->value, &representation_id, NULL, NULL))
415 representation_id = av_asprintf("%d", w->representation_id++);
416 if (!representation_id) return AVERROR(ENOMEM);
418 ret = write_representation(s, s->streams[as->streams[i]],
419 representation_id, !width_in_as,
420 !height_in_as, !sample_rate_in_as);
421 av_free(representation_id);
424 avio_printf(s->pb, "</AdaptationSet>\n");
428 static int to_integer(char *p, int len)
431 char *q = av_malloc(len);
433 return AVERROR(ENOMEM);
434 av_strlcpy(q, p, len);
440 static int parse_adaptation_sets(AVFormatContext *s)
442 WebMDashMuxContext *w = s->priv_data;
443 char *p = w->adaptation_sets;
445 enum { new_set, parsed_id, parsing_streams } state;
446 if (!w->adaptation_sets) {
447 av_log(s, AV_LOG_ERROR, "The 'adaptation_sets' option must be set.\n");
448 return AVERROR(EINVAL);
450 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
452 while (p < w->adaptation_sets + strlen(w->adaptation_sets)) {
455 else if (state == new_set && !strncmp(p, "id=", 3)) {
456 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
459 return AVERROR(ENOMEM);
462 w->as[w->nb_as - 1].nb_streams = 0;
463 w->as[w->nb_as - 1].streams = NULL;
464 p += 3; // consume "id="
465 q = w->as[w->nb_as - 1].id;
466 comma = strchr(p, ',');
467 if (!comma || comma - p >= sizeof(w->as[w->nb_as - 1].id)) {
468 av_log(s, AV_LOG_ERROR, "'id' in 'adaptation_sets' is malformed.\n");
469 return AVERROR(EINVAL);
471 while (*p != ',') *q++ = *p++;
475 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
476 p += 8; // consume "streams="
477 state = parsing_streams;
478 } else if (state == parsing_streams) {
479 struct AdaptationSet *as = &w->as[w->nb_as - 1];
480 int ret = av_reallocp_array(&as->streams, ++as->nb_streams,
481 sizeof(*as->streams));
485 while (*q != '\0' && *q != ',' && *q != ' ') q++;
486 as->streams[as->nb_streams - 1] = to_integer(p, q - p + 1);
487 if (as->streams[as->nb_streams - 1] < 0 ||
488 as->streams[as->nb_streams - 1] >= s->nb_streams) {
489 av_log(s, AV_LOG_ERROR, "Invalid value for 'streams' in adapation_sets.\n");
490 return AVERROR(EINVAL);
492 if (*q == '\0') break;
493 if (*q == ' ') state = new_set;
502 static int webm_dash_manifest_write_header(AVFormatContext *s)
507 WebMDashMuxContext *w = s->priv_data;
509 for (unsigned i = 0; i < s->nb_streams; i++) {
510 enum AVCodecID codec_id = s->streams[i]->codecpar->codec_id;
511 if (codec_id != AV_CODEC_ID_VP8 && codec_id != AV_CODEC_ID_VP9 &&
512 codec_id != AV_CODEC_ID_VORBIS && codec_id != AV_CODEC_ID_OPUS)
513 return AVERROR(EINVAL);
516 ret = parse_adaptation_sets(s);
520 ret = write_header(s);
524 avio_printf(s->pb, "<Period id=\"0\"");
525 avio_printf(s->pb, " start=\"PT%gS\"", start);
527 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
529 avio_printf(s->pb, " >\n");
531 for (i = 0; i < w->nb_as; i++) {
532 ret = write_adaptation_set(s, i);
538 avio_printf(s->pb, "</Period>\n");
541 free_adaptation_sets(s);
542 return ret < 0 ? ret : 0;
545 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
550 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
551 static const AVOption options[] = {
552 { "adaptation_sets", "Adaptation sets. Syntax: id=0,streams=0,1,2 id=1,streams=3,4 and so on", OFFSET(adaptation_sets), AV_OPT_TYPE_STRING, { 0 }, 0, 0, AV_OPT_FLAG_ENCODING_PARAM },
553 { "live", "create a live stream manifest", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
554 { "chunk_start_index", "start index of the chunk", OFFSET(chunk_start_index), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
555 { "chunk_duration_ms", "duration of each chunk (in milliseconds)", OFFSET(chunk_duration), AV_OPT_TYPE_INT, {.i64 = 1000}, 0, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
556 { "utc_timing_url", "URL of the page that will return the UTC timestamp in ISO format", OFFSET(utc_timing_url), AV_OPT_TYPE_STRING, { 0 }, 0, 0, AV_OPT_FLAG_ENCODING_PARAM },
557 { "time_shift_buffer_depth", "Smallest time (in seconds) shifting buffer for which any Representation is guaranteed to be available.", OFFSET(time_shift_buffer_depth), AV_OPT_TYPE_DOUBLE, { .dbl = 60.0 }, 1.0, DBL_MAX, AV_OPT_FLAG_ENCODING_PARAM },
558 { "minimum_update_period", "Minimum Update Period (in seconds) of the manifest.", OFFSET(minimum_update_period), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
562 static const AVClass webm_dash_class = {
563 .class_name = "WebM DASH Manifest muxer",
564 .item_name = av_default_item_name,
566 .version = LIBAVUTIL_VERSION_INT,
569 AVOutputFormat ff_webm_dash_manifest_muxer = {
570 .name = "webm_dash_manifest",
571 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
572 .mime_type = "application/xml",
574 .priv_data_size = sizeof(WebMDashMuxContext),
575 .write_header = webm_dash_manifest_write_header,
576 .write_packet = webm_dash_manifest_write_packet,
577 .priv_class = &webm_dash_class,