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
34 #include "avio_internal.h"
37 #include "libavutil/avstring.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/opt.h"
40 #include "libavutil/time_internal.h"
42 typedef struct AdaptationSet {
48 typedef struct WebMDashMuxContext {
50 char *adaptation_sets;
53 int representation_id;
55 int chunk_start_index;
58 double time_shift_buffer_depth;
59 int minimum_update_period;
62 static const char *get_codec_name(int codec_id)
64 return avcodec_descriptor_get(codec_id)->name;
67 static double get_duration(AVFormatContext *s)
71 for (i = 0; i < s->nb_streams; i++) {
72 AVDictionaryEntry *duration = av_dict_get(s->streams[i]->metadata,
74 if (!duration || atof(duration->value) < 0) continue;
75 if (atof(duration->value) > max) max = atof(duration->value);
80 static int write_header(AVFormatContext *s)
82 WebMDashMuxContext *w = s->priv_data;
83 double min_buffer_time = 1.0;
84 avio_printf(s->pb, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n");
85 avio_printf(s->pb, "<MPD\n");
86 avio_printf(s->pb, " xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n");
87 avio_printf(s->pb, " xmlns=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
88 avio_printf(s->pb, " xsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
89 avio_printf(s->pb, " type=\"%s\"\n", w->is_live ? "dynamic" : "static");
91 avio_printf(s->pb, " mediaPresentationDuration=\"PT%gS\"\n",
94 avio_printf(s->pb, " minBufferTime=\"PT%gS\"\n", min_buffer_time);
95 avio_printf(s->pb, " profiles=\"%s\"%s",
96 w->is_live ? "urn:mpeg:dash:profile:isoff-live:2011" : "urn:webm:dash:profile:webm-on-demand:2012",
97 w->is_live ? "\n" : ">\n");
99 time_t local_time = time(NULL);
100 struct tm gmt_buffer;
101 struct tm *gmt = gmtime_r(&local_time, &gmt_buffer);
103 if (!strftime(gmt_iso, 21, "%Y-%m-%dT%H:%M:%SZ", gmt)) {
104 return AVERROR_UNKNOWN;
106 if (s->flags & AVFMT_FLAG_BITEXACT) {
107 av_strlcpy(gmt_iso, "", 1);
109 avio_printf(s->pb, " availabilityStartTime=\"%s\"\n", gmt_iso);
110 avio_printf(s->pb, " timeShiftBufferDepth=\"PT%gS\"\n", w->time_shift_buffer_depth);
111 avio_printf(s->pb, " minimumUpdatePeriod=\"PT%dS\"", w->minimum_update_period);
112 avio_printf(s->pb, ">\n");
113 if (w->utc_timing_url) {
114 avio_printf(s->pb, "<UTCTiming\n");
115 avio_printf(s->pb, " schemeIdUri=\"urn:mpeg:dash:utc:http-iso:2014\"\n");
116 avio_printf(s->pb, " value=\"%s\"/>\n", w->utc_timing_url);
122 static void write_footer(AVFormatContext *s)
124 avio_printf(s->pb, "</MPD>\n");
127 static int subsegment_alignment(AVFormatContext *s, AdaptationSet *as) {
129 AVDictionaryEntry *gold = av_dict_get(s->streams[as->streams[0]]->metadata,
130 CUE_TIMESTAMPS, NULL, 0);
132 for (i = 1; i < as->nb_streams; i++) {
133 AVDictionaryEntry *ts = av_dict_get(s->streams[as->streams[i]]->metadata,
134 CUE_TIMESTAMPS, NULL, 0);
135 if (!ts || strncmp(gold->value, ts->value, strlen(gold->value))) return 0;
140 static int bitstream_switching(AVFormatContext *s, AdaptationSet *as) {
142 AVDictionaryEntry *gold_track_num = av_dict_get(s->streams[as->streams[0]]->metadata,
143 TRACK_NUMBER, NULL, 0);
144 AVCodecParameters *gold_par = s->streams[as->streams[0]]->codecpar;
145 if (!gold_track_num) return 0;
146 for (i = 1; i < as->nb_streams; i++) {
147 AVDictionaryEntry *track_num = av_dict_get(s->streams[as->streams[i]]->metadata,
148 TRACK_NUMBER, NULL, 0);
149 AVCodecParameters *par = s->streams[as->streams[i]]->codecpar;
151 strncmp(gold_track_num->value, track_num->value, strlen(gold_track_num->value)) ||
152 gold_par->codec_id != par->codec_id ||
153 gold_par->extradata_size != par->extradata_size ||
154 memcmp(gold_par->extradata, par->extradata, par->extradata_size)) {
162 * Writes a Representation within an Adaptation Set. Returns 0 on success and
165 static int write_representation(AVFormatContext *s, AVStream *stream, char *id,
166 int output_width, int output_height,
167 int output_sample_rate) {
168 WebMDashMuxContext *w = s->priv_data;
169 AVDictionaryEntry *irange = av_dict_get(stream->metadata, INITIALIZATION_RANGE, NULL, 0);
170 AVDictionaryEntry *cues_start = av_dict_get(stream->metadata, CUES_START, NULL, 0);
171 AVDictionaryEntry *cues_end = av_dict_get(stream->metadata, CUES_END, NULL, 0);
172 AVDictionaryEntry *filename = av_dict_get(stream->metadata, FILENAME, NULL, 0);
173 AVDictionaryEntry *bandwidth = av_dict_get(stream->metadata, BANDWIDTH, NULL, 0);
174 const char *bandwidth_str;
175 if ((w->is_live && (!filename)) ||
176 (!w->is_live && (!irange || !cues_start || !cues_end || !filename || !bandwidth))) {
177 return AVERROR_INVALIDDATA;
179 avio_printf(s->pb, "<Representation id=\"%s\"", id);
180 // if bandwidth for live was not provided, use a default
181 if (w->is_live && !bandwidth) {
182 bandwidth_str = (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) ? "128000" : "1000000";
184 bandwidth_str = bandwidth->value;
186 avio_printf(s->pb, " bandwidth=\"%s\"", bandwidth_str);
187 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
188 avio_printf(s->pb, " width=\"%d\"", stream->codecpar->width);
189 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
190 avio_printf(s->pb, " height=\"%d\"", stream->codecpar->height);
191 if (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO && output_sample_rate)
192 avio_printf(s->pb, " audioSamplingRate=\"%d\"", stream->codecpar->sample_rate);
194 // For live streams, Codec and Mime Type always go in the Representation tag.
195 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(stream->codecpar->codec_id));
196 avio_printf(s->pb, " mimeType=\"%s/webm\"",
197 stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
198 // For live streams, subsegments always start with key frames. So this
200 avio_printf(s->pb, " startsWithSAP=\"1\"");
201 avio_printf(s->pb, ">");
203 avio_printf(s->pb, ">\n");
204 avio_printf(s->pb, "<BaseURL>%s</BaseURL>\n", filename->value);
205 avio_printf(s->pb, "<SegmentBase\n");
206 avio_printf(s->pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
207 avio_printf(s->pb, "<Initialization\n");
208 avio_printf(s->pb, " range=\"0-%s\" />\n", irange->value);
209 avio_printf(s->pb, "</SegmentBase>\n");
211 avio_printf(s->pb, "</Representation>\n");
216 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
218 static int check_matching_width(AVFormatContext *s, AdaptationSet *as) {
220 if (as->nb_streams < 2) return 1;
221 first_width = s->streams[as->streams[0]]->codecpar->width;
222 for (i = 1; i < as->nb_streams; i++)
223 if (first_width != s->streams[as->streams[i]]->codecpar->width)
229 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
231 static int check_matching_height(AVFormatContext *s, AdaptationSet *as) {
233 if (as->nb_streams < 2) return 1;
234 first_height = s->streams[as->streams[0]]->codecpar->height;
235 for (i = 1; i < as->nb_streams; i++)
236 if (first_height != s->streams[as->streams[i]]->codecpar->height)
242 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
244 static int check_matching_sample_rate(AVFormatContext *s, AdaptationSet *as) {
245 int first_sample_rate, i;
246 if (as->nb_streams < 2) return 1;
247 first_sample_rate = s->streams[as->streams[0]]->codecpar->sample_rate;
248 for (i = 1; i < as->nb_streams; i++)
249 if (first_sample_rate != s->streams[as->streams[i]]->codecpar->sample_rate)
254 static void free_adaptation_sets(AVFormatContext *s) {
255 WebMDashMuxContext *w = s->priv_data;
257 for (i = 0; i < w->nb_as; i++) {
258 av_freep(&w->as[i].streams);
265 * Parses a live header filename and computes the representation id,
266 * initialization pattern and the media pattern. Pass NULL if you don't want to
267 * compute any of those 3. Returns 0 on success and non-zero on failure.
269 * Name of the header file should conform to the following pattern:
270 * <file_description>_<representation_id>.hdr where <file_description> can be
271 * anything. The chunks should be named according to the following pattern:
272 * <file_description>_<representation_id>_<chunk_number>.chk
274 static int parse_filename(char *filename, char **representation_id,
275 char **initialization_pattern, char **media_pattern) {
276 char *underscore_pos = NULL;
277 char *period_pos = NULL;
278 char *temp_pos = NULL;
279 char *filename_str = av_strdup(filename);
283 ret = AVERROR(ENOMEM);
286 temp_pos = av_stristr(filename_str, "_");
288 underscore_pos = temp_pos + 1;
289 temp_pos = av_stristr(temp_pos + 1, "_");
291 if (!underscore_pos) {
292 ret = AVERROR_INVALIDDATA;
295 period_pos = av_stristr(underscore_pos, ".");
297 ret = AVERROR_INVALIDDATA;
300 *(underscore_pos - 1) = 0;
301 if (representation_id) {
302 *representation_id = av_malloc(period_pos - underscore_pos + 1);
303 if (!(*representation_id)) {
304 ret = AVERROR(ENOMEM);
307 av_strlcpy(*representation_id, underscore_pos, period_pos - underscore_pos + 1);
309 if (initialization_pattern) {
310 *initialization_pattern = av_asprintf("%s_$RepresentationID$.hdr",
312 if (!(*initialization_pattern)) {
313 ret = AVERROR(ENOMEM);
318 *media_pattern = av_asprintf("%s_$RepresentationID$_$Number$.chk",
320 if (!(*media_pattern)) {
321 ret = AVERROR(ENOMEM);
327 av_freep(&filename_str);
332 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
334 static int write_adaptation_set(AVFormatContext *s, int as_index)
336 WebMDashMuxContext *w = s->priv_data;
337 AdaptationSet *as = &w->as[as_index];
338 AVCodecParameters *par = s->streams[as->streams[0]]->codecpar;
339 AVDictionaryEntry *lang;
341 static const char boolean[2][6] = { "false", "true" };
342 int subsegmentStartsWithSAP = 1;
344 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
345 // are the same for all contained Representations. otherwise, they will go
346 // on their respective Representation tag. For live streams, they always go
347 // in the Representation tag.
348 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
349 if (par->codec_type == AVMEDIA_TYPE_VIDEO) {
350 width_in_as = !w->is_live && check_matching_width(s, as);
351 height_in_as = !w->is_live && check_matching_height(s, as);
353 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
356 avio_printf(s->pb, "<AdaptationSet id=\"%s\"", as->id);
357 avio_printf(s->pb, " mimeType=\"%s/webm\"",
358 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
359 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(par->codec_id));
361 lang = av_dict_get(s->streams[as->streams[0]]->metadata, "language", NULL, 0);
362 if (lang) avio_printf(s->pb, " lang=\"%s\"", lang->value);
364 if (par->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
365 avio_printf(s->pb, " width=\"%d\"", par->width);
366 if (par->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
367 avio_printf(s->pb, " height=\"%d\"", par->height);
368 if (par->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
369 avio_printf(s->pb, " audioSamplingRate=\"%d\"", par->sample_rate);
371 avio_printf(s->pb, " bitstreamSwitching=\"%s\"",
372 boolean[bitstream_switching(s, as)]);
373 avio_printf(s->pb, " subsegmentAlignment=\"%s\"",
374 boolean[w->is_live || subsegment_alignment(s, as)]);
376 for (i = 0; i < as->nb_streams; i++) {
377 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
378 CLUSTER_KEYFRAME, NULL, 0);
379 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
381 avio_printf(s->pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
382 avio_printf(s->pb, ">\n");
385 AVDictionaryEntry *filename =
386 av_dict_get(s->streams[as->streams[0]]->metadata, FILENAME, NULL, 0);
387 char *initialization_pattern = NULL;
388 char *media_pattern = NULL;
389 int ret = parse_filename(filename->value, NULL, &initialization_pattern,
392 avio_printf(s->pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
393 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
394 avio_printf(s->pb, "<SegmentTemplate");
395 avio_printf(s->pb, " timescale=\"1000\"");
396 avio_printf(s->pb, " duration=\"%d\"", w->chunk_duration);
397 avio_printf(s->pb, " media=\"%s\"", media_pattern);
398 avio_printf(s->pb, " startNumber=\"%d\"", w->chunk_start_index);
399 avio_printf(s->pb, " initialization=\"%s\"", initialization_pattern);
400 avio_printf(s->pb, "/>\n");
401 av_free(initialization_pattern);
402 av_free(media_pattern);
405 for (i = 0; i < as->nb_streams; i++) {
406 char *representation_id = NULL;
409 AVDictionaryEntry *filename =
410 av_dict_get(s->streams[as->streams[i]]->metadata, FILENAME, NULL, 0);
412 return AVERROR(EINVAL);
413 if (ret = parse_filename(filename->value, &representation_id, NULL, NULL))
416 representation_id = av_asprintf("%d", w->representation_id++);
417 if (!representation_id) return AVERROR(ENOMEM);
419 ret = write_representation(s, s->streams[as->streams[i]],
420 representation_id, !width_in_as,
421 !height_in_as, !sample_rate_in_as);
422 av_free(representation_id);
425 avio_printf(s->pb, "</AdaptationSet>\n");
429 static int to_integer(char *p, int len)
432 char *q = av_malloc(len);
434 return AVERROR(ENOMEM);
435 av_strlcpy(q, p, len);
441 static int parse_adaptation_sets(AVFormatContext *s)
443 WebMDashMuxContext *w = s->priv_data;
444 char *p = w->adaptation_sets;
446 enum { new_set, parsed_id, parsing_streams } state;
447 if (!w->adaptation_sets) {
448 av_log(s, AV_LOG_ERROR, "The 'adaptation_sets' option must be set.\n");
449 return AVERROR(EINVAL);
451 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
453 while (p < w->adaptation_sets + strlen(w->adaptation_sets)) {
456 else if (state == new_set && !strncmp(p, "id=", 3)) {
457 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
460 return AVERROR(ENOMEM);
463 w->as[w->nb_as - 1].nb_streams = 0;
464 w->as[w->nb_as - 1].streams = NULL;
465 p += 3; // consume "id="
466 q = w->as[w->nb_as - 1].id;
467 comma = strchr(p, ',');
468 if (!comma || comma - p >= sizeof(w->as[w->nb_as - 1].id)) {
469 av_log(s, AV_LOG_ERROR, "'id' in 'adaptation_sets' is malformed.\n");
470 return AVERROR(EINVAL);
472 while (*p != ',') *q++ = *p++;
476 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
477 p += 8; // consume "streams="
478 state = parsing_streams;
479 } else if (state == parsing_streams) {
480 struct AdaptationSet *as = &w->as[w->nb_as - 1];
481 int ret = av_reallocp_array(&as->streams, ++as->nb_streams,
482 sizeof(*as->streams));
486 while (*q != '\0' && *q != ',' && *q != ' ') q++;
487 as->streams[as->nb_streams - 1] = to_integer(p, q - p + 1);
488 if (as->streams[as->nb_streams - 1] < 0 ||
489 as->streams[as->nb_streams - 1] >= s->nb_streams) {
490 av_log(s, AV_LOG_ERROR, "Invalid value for 'streams' in adapation_sets.\n");
491 return AVERROR(EINVAL);
493 if (*q == '\0') break;
494 if (*q == ' ') state = new_set;
503 static int webm_dash_manifest_write_header(AVFormatContext *s)
508 WebMDashMuxContext *w = s->priv_data;
510 for (unsigned i = 0; i < s->nb_streams; i++) {
511 enum AVCodecID codec_id = s->streams[i]->codecpar->codec_id;
512 if (codec_id != AV_CODEC_ID_VP8 && codec_id != AV_CODEC_ID_VP9 &&
513 codec_id != AV_CODEC_ID_VORBIS && codec_id != AV_CODEC_ID_OPUS)
514 return AVERROR(EINVAL);
517 ret = parse_adaptation_sets(s);
521 ret = write_header(s);
525 avio_printf(s->pb, "<Period id=\"0\"");
526 avio_printf(s->pb, " start=\"PT%gS\"", start);
528 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
530 avio_printf(s->pb, " >\n");
532 for (i = 0; i < w->nb_as; i++) {
533 ret = write_adaptation_set(s, i);
539 avio_printf(s->pb, "</Period>\n");
542 free_adaptation_sets(s);
543 return ret < 0 ? ret : 0;
546 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
551 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
552 static const AVOption options[] = {
553 { "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 },
554 { "live", "create a live stream manifest", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
555 { "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 },
556 { "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 },
557 { "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 },
558 { "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 },
559 { "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 },
563 #if CONFIG_WEBM_DASH_MANIFEST_MUXER
564 static const AVClass webm_dash_class = {
565 .class_name = "WebM DASH Manifest muxer",
566 .item_name = av_default_item_name,
568 .version = LIBAVUTIL_VERSION_INT,
571 AVOutputFormat ff_webm_dash_manifest_muxer = {
572 .name = "webm_dash_manifest",
573 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
574 .mime_type = "application/xml",
576 .priv_data_size = sizeof(WebMDashMuxContext),
577 .write_header = webm_dash_manifest_write_header,
578 .write_packet = webm_dash_manifest_write_packet,
579 .priv_class = &webm_dash_class,