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;
63 static const char *get_codec_name(int codec_id)
70 case AV_CODEC_ID_VORBIS:
72 case AV_CODEC_ID_OPUS:
78 static double get_duration(AVFormatContext *s)
82 for (i = 0; i < s->nb_streams; i++) {
83 AVDictionaryEntry *duration = av_dict_get(s->streams[i]->metadata,
85 if (!duration || atof(duration->value) < 0) continue;
86 if (atof(duration->value) > max) max = atof(duration->value);
91 static int write_header(AVFormatContext *s)
93 WebMDashMuxContext *w = s->priv_data;
94 double min_buffer_time = 1.0;
95 avio_printf(s->pb, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n");
96 avio_printf(s->pb, "<MPD\n");
97 avio_printf(s->pb, " xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n");
98 avio_printf(s->pb, " xmlns=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
99 avio_printf(s->pb, " xsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
100 avio_printf(s->pb, " type=\"%s\"\n", w->is_live ? "dynamic" : "static");
102 avio_printf(s->pb, " mediaPresentationDuration=\"PT%gS\"\n",
105 avio_printf(s->pb, " minBufferTime=\"PT%gS\"\n", min_buffer_time);
106 avio_printf(s->pb, " profiles=\"%s\"%s",
107 w->is_live ? "urn:mpeg:dash:profile:isoff-live:2011" : "urn:webm:dash:profile:webm-on-demand:2012",
108 w->is_live ? "\n" : ">\n");
110 time_t local_time = time(NULL);
111 struct tm gmt_buffer;
112 struct tm *gmt = gmtime_r(&local_time, &gmt_buffer);
114 if (!strftime(gmt_iso, 21, "%Y-%m-%dT%H:%M:%SZ", gmt)) {
115 return AVERROR_UNKNOWN;
118 av_strlcpy(gmt_iso, "", 1);
120 avio_printf(s->pb, " availabilityStartTime=\"%s\"\n", gmt_iso);
121 avio_printf(s->pb, " timeShiftBufferDepth=\"PT%gS\"\n", w->time_shift_buffer_depth);
122 avio_printf(s->pb, " minimumUpdatePeriod=\"PT%dS\"", w->minimum_update_period);
123 avio_printf(s->pb, ">\n");
124 if (w->utc_timing_url) {
125 avio_printf(s->pb, "<UTCTiming\n");
126 avio_printf(s->pb, " schemeIdUri=\"urn:mpeg:dash:utc:http-iso:2014\"\n");
127 avio_printf(s->pb, " value=\"%s\"/>\n", w->utc_timing_url);
133 static void write_footer(AVFormatContext *s)
135 avio_printf(s->pb, "</MPD>\n");
138 static int subsegment_alignment(AVFormatContext *s, AdaptationSet *as) {
140 AVDictionaryEntry *gold = av_dict_get(s->streams[as->streams[0]]->metadata,
141 CUE_TIMESTAMPS, NULL, 0);
143 for (i = 1; i < as->nb_streams; i++) {
144 AVDictionaryEntry *ts = av_dict_get(s->streams[as->streams[i]]->metadata,
145 CUE_TIMESTAMPS, NULL, 0);
146 if (!ts || strncmp(gold->value, ts->value, strlen(gold->value))) return 0;
151 static int bitstream_switching(AVFormatContext *s, AdaptationSet *as) {
153 AVDictionaryEntry *gold_track_num = av_dict_get(s->streams[as->streams[0]]->metadata,
154 TRACK_NUMBER, NULL, 0);
155 AVCodecParameters *gold_par = s->streams[as->streams[0]]->codecpar;
156 if (!gold_track_num) return 0;
157 for (i = 1; i < as->nb_streams; i++) {
158 AVDictionaryEntry *track_num = av_dict_get(s->streams[as->streams[i]]->metadata,
159 TRACK_NUMBER, NULL, 0);
160 AVCodecParameters *par = s->streams[as->streams[i]]->codecpar;
162 strncmp(gold_track_num->value, track_num->value, strlen(gold_track_num->value)) ||
163 gold_par->codec_id != par->codec_id ||
164 gold_par->extradata_size != par->extradata_size ||
165 memcmp(gold_par->extradata, par->extradata, par->extradata_size)) {
173 * Writes a Representation within an Adaptation Set. Returns 0 on success and
176 static int write_representation(AVFormatContext *s, AVStream *stream, char *id,
177 int output_width, int output_height,
178 int output_sample_rate) {
179 WebMDashMuxContext *w = s->priv_data;
180 AVDictionaryEntry *irange = av_dict_get(stream->metadata, INITIALIZATION_RANGE, NULL, 0);
181 AVDictionaryEntry *cues_start = av_dict_get(stream->metadata, CUES_START, NULL, 0);
182 AVDictionaryEntry *cues_end = av_dict_get(stream->metadata, CUES_END, NULL, 0);
183 AVDictionaryEntry *filename = av_dict_get(stream->metadata, FILENAME, NULL, 0);
184 AVDictionaryEntry *bandwidth = av_dict_get(stream->metadata, BANDWIDTH, NULL, 0);
185 const char *bandwidth_str;
186 if ((w->is_live && (!filename)) ||
187 (!w->is_live && (!irange || !cues_start || !cues_end || !filename || !bandwidth))) {
188 return AVERROR_INVALIDDATA;
190 avio_printf(s->pb, "<Representation id=\"%s\"", id);
191 // if bandwidth for live was not provided, use a default
192 if (w->is_live && !bandwidth) {
193 bandwidth_str = (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) ? "128000" : "1000000";
195 bandwidth_str = bandwidth->value;
197 avio_printf(s->pb, " bandwidth=\"%s\"", bandwidth_str);
198 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
199 avio_printf(s->pb, " width=\"%d\"", stream->codecpar->width);
200 if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
201 avio_printf(s->pb, " height=\"%d\"", stream->codecpar->height);
202 if (stream->codecpar->codec_type == AVMEDIA_TYPE_AUDIO && output_sample_rate)
203 avio_printf(s->pb, " audioSamplingRate=\"%d\"", stream->codecpar->sample_rate);
205 // For live streams, Codec and Mime Type always go in the Representation tag.
206 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(stream->codecpar->codec_id));
207 avio_printf(s->pb, " mimeType=\"%s/webm\"",
208 stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
209 // For live streams, subsegments always start with key frames. So this
211 avio_printf(s->pb, " startsWithSAP=\"1\"");
212 avio_printf(s->pb, ">");
214 avio_printf(s->pb, ">\n");
215 avio_printf(s->pb, "<BaseURL>%s</BaseURL>\n", filename->value);
216 avio_printf(s->pb, "<SegmentBase\n");
217 avio_printf(s->pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
218 avio_printf(s->pb, "<Initialization\n");
219 avio_printf(s->pb, " range=\"0-%s\" />\n", irange->value);
220 avio_printf(s->pb, "</SegmentBase>\n");
222 avio_printf(s->pb, "</Representation>\n");
227 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
229 static int check_matching_width(AVFormatContext *s, AdaptationSet *as) {
231 if (as->nb_streams < 2) return 1;
232 first_width = s->streams[as->streams[0]]->codecpar->width;
233 for (i = 1; i < as->nb_streams; i++)
234 if (first_width != s->streams[as->streams[i]]->codecpar->width)
240 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
242 static int check_matching_height(AVFormatContext *s, AdaptationSet *as) {
244 if (as->nb_streams < 2) return 1;
245 first_height = s->streams[as->streams[0]]->codecpar->height;
246 for (i = 1; i < as->nb_streams; i++)
247 if (first_height != s->streams[as->streams[i]]->codecpar->height)
253 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
255 static int check_matching_sample_rate(AVFormatContext *s, AdaptationSet *as) {
256 int first_sample_rate, i;
257 if (as->nb_streams < 2) return 1;
258 first_sample_rate = s->streams[as->streams[0]]->codecpar->sample_rate;
259 for (i = 1; i < as->nb_streams; i++)
260 if (first_sample_rate != s->streams[as->streams[i]]->codecpar->sample_rate)
265 static void free_adaptation_sets(AVFormatContext *s) {
266 WebMDashMuxContext *w = s->priv_data;
268 for (i = 0; i < w->nb_as; i++) {
269 av_freep(&w->as[i].streams);
276 * Parses a live header filename and computes the representation id,
277 * initialization pattern and the media pattern. Pass NULL if you don't want to
278 * compute any of those 3. Returns 0 on success and non-zero on failure.
280 * Name of the header file should conform to the following pattern:
281 * <file_description>_<representation_id>.hdr where <file_description> can be
282 * anything. The chunks should be named according to the following pattern:
283 * <file_description>_<representation_id>_<chunk_number>.chk
285 static int parse_filename(char *filename, char **representation_id,
286 char **initialization_pattern, char **media_pattern) {
287 char *underscore_pos = NULL;
288 char *period_pos = NULL;
289 char *temp_pos = NULL;
290 char *filename_str = av_strdup(filename);
294 ret = AVERROR(ENOMEM);
297 temp_pos = av_stristr(filename_str, "_");
299 underscore_pos = temp_pos + 1;
300 temp_pos = av_stristr(temp_pos + 1, "_");
302 if (!underscore_pos) {
303 ret = AVERROR_INVALIDDATA;
306 period_pos = av_stristr(underscore_pos, ".");
308 ret = AVERROR_INVALIDDATA;
311 *(underscore_pos - 1) = 0;
312 if (representation_id) {
313 *representation_id = av_malloc(period_pos - underscore_pos + 1);
314 if (!(*representation_id)) {
315 ret = AVERROR(ENOMEM);
318 av_strlcpy(*representation_id, underscore_pos, period_pos - underscore_pos + 1);
320 if (initialization_pattern) {
321 *initialization_pattern = av_asprintf("%s_$RepresentationID$.hdr",
323 if (!(*initialization_pattern)) {
324 ret = AVERROR(ENOMEM);
329 *media_pattern = av_asprintf("%s_$RepresentationID$_$Number$.chk",
331 if (!(*media_pattern)) {
332 ret = AVERROR(ENOMEM);
338 av_freep(&filename_str);
343 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
345 static int write_adaptation_set(AVFormatContext *s, int as_index)
347 WebMDashMuxContext *w = s->priv_data;
348 AdaptationSet *as = &w->as[as_index];
349 AVCodecParameters *par = s->streams[as->streams[0]]->codecpar;
350 AVDictionaryEntry *lang;
352 static const char boolean[2][6] = { "false", "true" };
353 int subsegmentStartsWithSAP = 1;
355 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
356 // are the same for all contained Representations. otherwise, they will go
357 // on their respective Representation tag. For live streams, they always go
358 // in the Representation tag.
359 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
360 if (par->codec_type == AVMEDIA_TYPE_VIDEO) {
361 width_in_as = !w->is_live && check_matching_width(s, as);
362 height_in_as = !w->is_live && check_matching_height(s, as);
364 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
367 avio_printf(s->pb, "<AdaptationSet id=\"%s\"", as->id);
368 avio_printf(s->pb, " mimeType=\"%s/webm\"",
369 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
370 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(par->codec_id));
372 lang = av_dict_get(s->streams[as->streams[0]]->metadata, "language", NULL, 0);
373 if (lang) avio_printf(s->pb, " lang=\"%s\"", lang->value);
375 if (par->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
376 avio_printf(s->pb, " width=\"%d\"", par->width);
377 if (par->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
378 avio_printf(s->pb, " height=\"%d\"", par->height);
379 if (par->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
380 avio_printf(s->pb, " audioSamplingRate=\"%d\"", par->sample_rate);
382 avio_printf(s->pb, " bitstreamSwitching=\"%s\"",
383 boolean[bitstream_switching(s, as)]);
384 avio_printf(s->pb, " subsegmentAlignment=\"%s\"",
385 boolean[w->is_live || subsegment_alignment(s, as)]);
387 for (i = 0; i < as->nb_streams; i++) {
388 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
389 CLUSTER_KEYFRAME, NULL, 0);
390 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
392 avio_printf(s->pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
393 avio_printf(s->pb, ">\n");
396 AVDictionaryEntry *filename =
397 av_dict_get(s->streams[as->streams[0]]->metadata, FILENAME, NULL, 0);
398 char *initialization_pattern = NULL;
399 char *media_pattern = NULL;
400 int ret = parse_filename(filename->value, NULL, &initialization_pattern,
403 avio_printf(s->pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
404 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
405 avio_printf(s->pb, "<SegmentTemplate");
406 avio_printf(s->pb, " timescale=\"1000\"");
407 avio_printf(s->pb, " duration=\"%d\"", w->chunk_duration);
408 avio_printf(s->pb, " media=\"%s\"", media_pattern);
409 avio_printf(s->pb, " startNumber=\"%d\"", w->chunk_start_index);
410 avio_printf(s->pb, " initialization=\"%s\"", initialization_pattern);
411 avio_printf(s->pb, "/>\n");
412 av_free(initialization_pattern);
413 av_free(media_pattern);
416 for (i = 0; i < as->nb_streams; i++) {
417 char *representation_id = NULL;
420 AVDictionaryEntry *filename =
421 av_dict_get(s->streams[as->streams[i]]->metadata, FILENAME, NULL, 0);
423 return AVERROR(EINVAL);
424 if (ret = parse_filename(filename->value, &representation_id, NULL, NULL))
427 representation_id = av_asprintf("%d", w->representation_id++);
428 if (!representation_id) return AVERROR(ENOMEM);
430 ret = write_representation(s, s->streams[as->streams[i]],
431 representation_id, !width_in_as,
432 !height_in_as, !sample_rate_in_as);
433 av_free(representation_id);
436 avio_printf(s->pb, "</AdaptationSet>\n");
440 static int to_integer(char *p, int len)
443 char *q = av_malloc(sizeof(char) * len);
445 return AVERROR(ENOMEM);
446 av_strlcpy(q, p, len);
452 static int parse_adaptation_sets(AVFormatContext *s)
454 WebMDashMuxContext *w = s->priv_data;
455 char *p = w->adaptation_sets;
457 enum { new_set, parsed_id, parsing_streams } state;
458 if (!w->adaptation_sets) {
459 av_log(s, AV_LOG_ERROR, "The 'adaptation_sets' option must be set.\n");
460 return AVERROR(EINVAL);
462 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
464 while (p < w->adaptation_sets + strlen(w->adaptation_sets)) {
467 else if (state == new_set && !strncmp(p, "id=", 3)) {
468 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
470 return AVERROR(ENOMEM);
473 w->as[w->nb_as - 1].nb_streams = 0;
474 w->as[w->nb_as - 1].streams = NULL;
475 p += 3; // consume "id="
476 q = w->as[w->nb_as - 1].id;
477 while (*p != ',') *q++ = *p++;
481 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
482 p += 8; // consume "streams="
483 state = parsing_streams;
484 } else if (state == parsing_streams) {
485 struct AdaptationSet *as = &w->as[w->nb_as - 1];
487 while (*q != '\0' && *q != ',' && *q != ' ') q++;
488 as->streams = av_realloc(as->streams, sizeof(*as->streams) * ++as->nb_streams);
489 if (as->streams == NULL)
490 return AVERROR(ENOMEM);
491 as->streams[as->nb_streams - 1] = to_integer(p, q - p + 1);
492 if (as->streams[as->nb_streams - 1] < 0 ||
493 as->streams[as->nb_streams - 1] >= s->nb_streams) {
494 av_log(s, AV_LOG_ERROR, "Invalid value for 'streams' in adapation_sets.\n");
495 return AVERROR(EINVAL);
497 if (*q == '\0') break;
498 if (*q == ' ') state = new_set;
507 static int webm_dash_manifest_write_header(AVFormatContext *s)
512 WebMDashMuxContext *w = s->priv_data;
513 ret = parse_adaptation_sets(s);
517 ret = write_header(s);
521 avio_printf(s->pb, "<Period id=\"0\"");
522 avio_printf(s->pb, " start=\"PT%gS\"", start);
524 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
526 avio_printf(s->pb, " >\n");
528 for (i = 0; i < w->nb_as; i++) {
529 ret = write_adaptation_set(s, i);
535 avio_printf(s->pb, "</Period>\n");
538 free_adaptation_sets(s);
539 return ret < 0 ? ret : 0;
542 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
547 static int webm_dash_manifest_write_trailer(AVFormatContext *s)
549 free_adaptation_sets(s);
553 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
554 static const AVOption options[] = {
555 { "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 },
556 { "debug_mode", "[private option - users should never set this]. Create deterministic output", OFFSET(debug_mode), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
557 { "live", "create a live stream manifest", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
558 { "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 },
559 { "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 },
560 { "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 },
561 { "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 },
562 { "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 },
566 #if CONFIG_WEBM_DASH_MANIFEST_MUXER
567 static const AVClass webm_dash_class = {
568 .class_name = "WebM DASH Manifest muxer",
569 .item_name = av_default_item_name,
571 .version = LIBAVUTIL_VERSION_INT,
574 AVOutputFormat ff_webm_dash_manifest_muxer = {
575 .name = "webm_dash_manifest",
576 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
577 .mime_type = "application/xml",
579 .priv_data_size = sizeof(WebMDashMuxContext),
580 .write_header = webm_dash_manifest_write_header,
581 .write_packet = webm_dash_manifest_write_packet,
582 .write_trailer = webm_dash_manifest_write_trailer,
583 .priv_class = &webm_dash_class,