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 void 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 strftime(gmt_iso, 21, "%Y-%m-%dT%H:%M:%SZ", gmt);
116 av_strlcpy(gmt_iso, "", 1);
118 avio_printf(s->pb, " availabilityStartTime=\"%s\"\n", gmt_iso);
119 avio_printf(s->pb, " timeShiftBufferDepth=\"PT%gS\"\n", w->time_shift_buffer_depth);
120 avio_printf(s->pb, " minimumUpdatePeriod=\"PT%dS\"", w->minimum_update_period);
121 avio_printf(s->pb, ">\n");
122 if (w->utc_timing_url) {
123 avio_printf(s->pb, "<UTCTiming\n");
124 avio_printf(s->pb, " schemeIdUri=\"urn:mpeg:dash:utc:http-iso:2014\"\n");
125 avio_printf(s->pb, " value=\"%s\"/>\n", w->utc_timing_url);
130 static void write_footer(AVFormatContext *s)
132 avio_printf(s->pb, "</MPD>\n");
135 static int subsegment_alignment(AVFormatContext *s, AdaptationSet *as) {
137 AVDictionaryEntry *gold = av_dict_get(s->streams[as->streams[0]]->metadata,
138 CUE_TIMESTAMPS, NULL, 0);
140 for (i = 1; i < as->nb_streams; i++) {
141 AVDictionaryEntry *ts = av_dict_get(s->streams[as->streams[i]]->metadata,
142 CUE_TIMESTAMPS, NULL, 0);
143 if (!ts || strncmp(gold->value, ts->value, strlen(gold->value))) return 0;
148 static int bitstream_switching(AVFormatContext *s, AdaptationSet *as) {
150 AVDictionaryEntry *gold_track_num = av_dict_get(s->streams[as->streams[0]]->metadata,
151 TRACK_NUMBER, NULL, 0);
152 AVCodecContext *gold_codec = s->streams[as->streams[0]]->codec;
153 if (!gold_track_num) return 0;
154 for (i = 1; i < as->nb_streams; i++) {
155 AVDictionaryEntry *track_num = av_dict_get(s->streams[as->streams[i]]->metadata,
156 TRACK_NUMBER, NULL, 0);
157 AVCodecContext *codec = s->streams[as->streams[i]]->codec;
159 strncmp(gold_track_num->value, track_num->value, strlen(gold_track_num->value)) ||
160 gold_codec->codec_id != codec->codec_id ||
161 gold_codec->extradata_size != codec->extradata_size ||
162 memcmp(gold_codec->extradata, codec->extradata, codec->extradata_size)) {
170 * Writes a Representation within an Adaptation Set. Returns 0 on success and
173 static int write_representation(AVFormatContext *s, AVStream *stream, char *id,
174 int output_width, int output_height,
175 int output_sample_rate) {
176 WebMDashMuxContext *w = s->priv_data;
177 AVDictionaryEntry *irange = av_dict_get(stream->metadata, INITIALIZATION_RANGE, NULL, 0);
178 AVDictionaryEntry *cues_start = av_dict_get(stream->metadata, CUES_START, NULL, 0);
179 AVDictionaryEntry *cues_end = av_dict_get(stream->metadata, CUES_END, NULL, 0);
180 AVDictionaryEntry *filename = av_dict_get(stream->metadata, FILENAME, NULL, 0);
181 AVDictionaryEntry *bandwidth = av_dict_get(stream->metadata, BANDWIDTH, NULL, 0);
182 if ((w->is_live && (!filename)) ||
183 (!w->is_live && (!irange || !cues_start || !cues_end || !filename || !bandwidth))) {
184 return AVERROR_INVALIDDATA;
186 avio_printf(s->pb, "<Representation id=\"%s\"", id);
187 // FIXME: For live, This should be obtained from the input file or as an AVOption.
188 avio_printf(s->pb, " bandwidth=\"%s\"",
189 w->is_live ? (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO ? "128000" : "1000000") : bandwidth->value);
190 if (stream->codec->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
191 avio_printf(s->pb, " width=\"%d\"", stream->codec->width);
192 if (stream->codec->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
193 avio_printf(s->pb, " height=\"%d\"", stream->codec->height);
194 if (stream->codec->codec_type = AVMEDIA_TYPE_AUDIO && output_sample_rate)
195 avio_printf(s->pb, " audioSamplingRate=\"%d\"", stream->codec->sample_rate);
197 // For live streams, Codec and Mime Type always go in the Representation tag.
198 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(stream->codec->codec_id));
199 avio_printf(s->pb, " mimeType=\"%s/webm\"",
200 stream->codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
201 // For live streams, subsegments always start with key frames. So this
203 avio_printf(s->pb, " startsWithSAP=\"1\"");
204 avio_printf(s->pb, ">");
206 avio_printf(s->pb, ">\n");
207 avio_printf(s->pb, "<BaseURL>%s</BaseURL>\n", filename->value);
208 avio_printf(s->pb, "<SegmentBase\n");
209 avio_printf(s->pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
210 avio_printf(s->pb, "<Initialization\n");
211 avio_printf(s->pb, " range=\"0-%s\" />\n", irange->value);
212 avio_printf(s->pb, "</SegmentBase>\n");
214 avio_printf(s->pb, "</Representation>\n");
219 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
221 static int check_matching_width(AVFormatContext *s, AdaptationSet *as) {
223 if (as->nb_streams < 2) return 1;
224 first_width = s->streams[as->streams[0]]->codec->width;
225 for (i = 1; i < as->nb_streams; i++)
226 if (first_width != s->streams[as->streams[i]]->codec->width)
232 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
234 static int check_matching_height(AVFormatContext *s, AdaptationSet *as) {
236 if (as->nb_streams < 2) return 1;
237 first_height = s->streams[as->streams[0]]->codec->height;
238 for (i = 1; i < as->nb_streams; i++)
239 if (first_height != s->streams[as->streams[i]]->codec->height)
245 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
247 static int check_matching_sample_rate(AVFormatContext *s, AdaptationSet *as) {
248 int first_sample_rate, i;
249 if (as->nb_streams < 2) return 1;
250 first_sample_rate = s->streams[as->streams[0]]->codec->sample_rate;
251 for (i = 1; i < as->nb_streams; i++)
252 if (first_sample_rate != s->streams[as->streams[i]]->codec->sample_rate)
257 static void free_adaptation_sets(AVFormatContext *s) {
258 WebMDashMuxContext *w = s->priv_data;
260 for (i = 0; i < w->nb_as; i++) {
261 av_freep(&w->as[i].streams);
268 * Parses a live header filename and computes the representation id,
269 * initialization pattern and the media pattern. Pass NULL if you don't want to
270 * compute any of those 3. Returns 0 on success and non-zero on failure.
272 * Name of the header file should conform to the following pattern:
273 * <file_description>_<representation_id>.hdr where <file_description> can be
274 * anything. The chunks should be named according to the following pattern:
275 * <file_description>_<representation_id>_<chunk_number>.chk
277 static int parse_filename(char *filename, char **representation_id,
278 char **initialization_pattern, char **media_pattern) {
279 char *underscore_pos = NULL;
280 char *period_pos = NULL;
281 char *temp_pos = NULL;
282 char *filename_str = av_strdup(filename);
283 if (!filename_str) return AVERROR(ENOMEM);
284 temp_pos = av_stristr(filename_str, "_");
286 underscore_pos = temp_pos + 1;
287 temp_pos = av_stristr(temp_pos + 1, "_");
289 if (!underscore_pos) return AVERROR_INVALIDDATA;
290 period_pos = av_stristr(underscore_pos, ".");
291 if (!period_pos) return AVERROR_INVALIDDATA;
292 *(underscore_pos - 1) = 0;
293 if (representation_id) {
294 *representation_id = av_malloc(period_pos - underscore_pos + 1);
295 if (!(*representation_id)) return AVERROR(ENOMEM);
296 av_strlcpy(*representation_id, underscore_pos, period_pos - underscore_pos + 1);
298 if (initialization_pattern) {
299 *initialization_pattern = av_asprintf("%s_$RepresentationID$.hdr",
301 if (!(*initialization_pattern)) return AVERROR(ENOMEM);
304 *media_pattern = av_asprintf("%s_$RepresentationID$_$Number$.chk",
306 if (!(*media_pattern)) return AVERROR(ENOMEM);
308 av_free(filename_str);
313 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
315 static int write_adaptation_set(AVFormatContext *s, int as_index)
317 WebMDashMuxContext *w = s->priv_data;
318 AdaptationSet *as = &w->as[as_index];
319 AVCodecContext *codec = s->streams[as->streams[0]]->codec;
320 AVDictionaryEntry *lang;
322 static const char boolean[2][6] = { "false", "true" };
323 int subsegmentStartsWithSAP = 1;
325 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
326 // are the same for all contained Representations. otherwise, they will go
327 // on their respective Representation tag. For live streams, they always go
328 // in the Representation tag.
329 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
330 if (codec->codec_type == AVMEDIA_TYPE_VIDEO) {
331 width_in_as = !w->is_live && check_matching_width(s, as);
332 height_in_as = !w->is_live && check_matching_height(s, as);
334 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
337 avio_printf(s->pb, "<AdaptationSet id=\"%s\"", as->id);
338 avio_printf(s->pb, " mimeType=\"%s/webm\"",
339 codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
340 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(codec->codec_id));
342 lang = av_dict_get(s->streams[as->streams[0]]->metadata, "language", NULL, 0);
343 if (lang) avio_printf(s->pb, " lang=\"%s\"", lang->value);
345 if (codec->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
346 avio_printf(s->pb, " width=\"%d\"", codec->width);
347 if (codec->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
348 avio_printf(s->pb, " height=\"%d\"", codec->height);
349 if (codec->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
350 avio_printf(s->pb, " audioSamplingRate=\"%d\"", codec->sample_rate);
352 avio_printf(s->pb, " bitstreamSwitching=\"%s\"",
353 boolean[bitstream_switching(s, as)]);
354 avio_printf(s->pb, " subsegmentAlignment=\"%s\"",
355 boolean[w->is_live || subsegment_alignment(s, as)]);
357 for (i = 0; i < as->nb_streams; i++) {
358 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
359 CLUSTER_KEYFRAME, NULL, 0);
360 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
362 avio_printf(s->pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
363 avio_printf(s->pb, ">\n");
366 AVDictionaryEntry *filename =
367 av_dict_get(s->streams[as->streams[0]]->metadata, FILENAME, NULL, 0);
368 char *initialization_pattern = NULL;
369 char *media_pattern = NULL;
370 int ret = parse_filename(filename->value, NULL, &initialization_pattern,
373 avio_printf(s->pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
374 codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
375 avio_printf(s->pb, "<SegmentTemplate");
376 avio_printf(s->pb, " timescale=\"1000\"");
377 avio_printf(s->pb, " duration=\"%d\"", w->chunk_duration);
378 avio_printf(s->pb, " media=\"%s\"", media_pattern);
379 avio_printf(s->pb, " startNumber=\"%d\"", w->chunk_start_index);
380 avio_printf(s->pb, " initialization=\"%s\"", initialization_pattern);
381 avio_printf(s->pb, "/>\n");
382 av_free(initialization_pattern);
383 av_free(media_pattern);
386 for (i = 0; i < as->nb_streams; i++) {
387 char *representation_id = NULL;
390 AVDictionaryEntry *filename =
391 av_dict_get(s->streams[as->streams[i]]->metadata, FILENAME, NULL, 0);
393 (ret = parse_filename(filename->value, &representation_id, NULL, NULL))) {
397 representation_id = av_asprintf("%d", w->representation_id++);
398 if (!representation_id) return AVERROR(ENOMEM);
400 ret = write_representation(s, s->streams[as->streams[i]],
401 representation_id, !width_in_as,
402 !height_in_as, !sample_rate_in_as);
403 av_free(representation_id);
406 avio_printf(s->pb, "</AdaptationSet>\n");
410 static int to_integer(char *p, int len)
413 char *q = av_malloc(sizeof(char) * len);
415 return AVERROR(ENOMEM);
416 av_strlcpy(q, p, len);
422 static int parse_adaptation_sets(AVFormatContext *s)
424 WebMDashMuxContext *w = s->priv_data;
425 char *p = w->adaptation_sets;
427 enum { new_set, parsed_id, parsing_streams } state;
428 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
430 while (p < w->adaptation_sets + strlen(w->adaptation_sets)) {
433 else if (state == new_set && !strncmp(p, "id=", 3)) {
434 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
436 return AVERROR(ENOMEM);
439 w->as[w->nb_as - 1].nb_streams = 0;
440 w->as[w->nb_as - 1].streams = NULL;
441 p += 3; // consume "id="
442 q = w->as[w->nb_as - 1].id;
443 while (*p != ',') *q++ = *p++;
447 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
448 p += 8; // consume "streams="
449 state = parsing_streams;
450 } else if (state == parsing_streams) {
451 struct AdaptationSet *as = &w->as[w->nb_as - 1];
453 while (*q != '\0' && *q != ',' && *q != ' ') q++;
454 as->streams = av_realloc(as->streams, sizeof(*as->streams) * ++as->nb_streams);
455 if (as->streams == NULL)
456 return AVERROR(ENOMEM);
457 as->streams[as->nb_streams - 1] = to_integer(p, q - p + 1);
458 if (as->streams[as->nb_streams - 1] < 0) return -1;
459 if (*q == '\0') break;
460 if (*q == ' ') state = new_set;
469 static int webm_dash_manifest_write_header(AVFormatContext *s)
474 WebMDashMuxContext *w = s->priv_data;
475 ret = parse_adaptation_sets(s);
477 free_adaptation_sets(s);
481 avio_printf(s->pb, "<Period id=\"0\"");
482 avio_printf(s->pb, " start=\"PT%gS\"", start);
484 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
486 avio_printf(s->pb, " >\n");
488 for (i = 0; i < w->nb_as; i++) {
489 ret = write_adaptation_set(s, i);
491 free_adaptation_sets(s);
496 avio_printf(s->pb, "</Period>\n");
501 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
506 static int webm_dash_manifest_write_trailer(AVFormatContext *s)
508 free_adaptation_sets(s);
512 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
513 static const AVOption options[] = {
514 { "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 },
515 { "debug_mode", "[private option - users should never set this]. set this to 1 to create deterministic output", OFFSET(debug_mode), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
516 { "live", "set this to 1 to create a live stream manifest", OFFSET(is_live), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
517 { "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 },
518 { "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 },
519 { "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 },
520 { "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 },
521 { "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 },
525 #if CONFIG_WEBM_DASH_MANIFEST_MUXER
526 static const AVClass webm_dash_class = {
527 .class_name = "WebM DASH Manifest muxer",
528 .item_name = av_default_item_name,
530 .version = LIBAVUTIL_VERSION_INT,
533 AVOutputFormat ff_webm_dash_manifest_muxer = {
534 .name = "webm_dash_manifest",
535 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
536 .mime_type = "application/xml",
538 .priv_data_size = sizeof(WebMDashMuxContext),
539 .write_header = webm_dash_manifest_write_header,
540 .write_packet = webm_dash_manifest_write_packet,
541 .write_trailer = webm_dash_manifest_write_trailer,
542 .priv_class = &webm_dash_class,