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 AVCodecContext *gold_codec = s->streams[as->streams[0]]->codec;
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 AVCodecContext *codec = s->streams[as->streams[i]]->codec;
162 strncmp(gold_track_num->value, track_num->value, strlen(gold_track_num->value)) ||
163 gold_codec->codec_id != codec->codec_id ||
164 gold_codec->extradata_size != codec->extradata_size ||
165 memcmp(gold_codec->extradata, codec->extradata, codec->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 if ((w->is_live && (!filename)) ||
186 (!w->is_live && (!irange || !cues_start || !cues_end || !filename || !bandwidth))) {
187 return AVERROR_INVALIDDATA;
189 avio_printf(s->pb, "<Representation id=\"%s\"", id);
190 // FIXME: For live, This should be obtained from the input file or as an AVOption.
191 avio_printf(s->pb, " bandwidth=\"%s\"",
192 w->is_live ? (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO ? "128000" : "1000000") : bandwidth->value);
193 if (stream->codec->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
194 avio_printf(s->pb, " width=\"%d\"", stream->codec->width);
195 if (stream->codec->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
196 avio_printf(s->pb, " height=\"%d\"", stream->codec->height);
197 if (stream->codec->codec_type == AVMEDIA_TYPE_AUDIO && output_sample_rate)
198 avio_printf(s->pb, " audioSamplingRate=\"%d\"", stream->codec->sample_rate);
200 // For live streams, Codec and Mime Type always go in the Representation tag.
201 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(stream->codec->codec_id));
202 avio_printf(s->pb, " mimeType=\"%s/webm\"",
203 stream->codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
204 // For live streams, subsegments always start with key frames. So this
206 avio_printf(s->pb, " startsWithSAP=\"1\"");
207 avio_printf(s->pb, ">");
209 avio_printf(s->pb, ">\n");
210 avio_printf(s->pb, "<BaseURL>%s</BaseURL>\n", filename->value);
211 avio_printf(s->pb, "<SegmentBase\n");
212 avio_printf(s->pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
213 avio_printf(s->pb, "<Initialization\n");
214 avio_printf(s->pb, " range=\"0-%s\" />\n", irange->value);
215 avio_printf(s->pb, "</SegmentBase>\n");
217 avio_printf(s->pb, "</Representation>\n");
222 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
224 static int check_matching_width(AVFormatContext *s, AdaptationSet *as) {
226 if (as->nb_streams < 2) return 1;
227 first_width = s->streams[as->streams[0]]->codec->width;
228 for (i = 1; i < as->nb_streams; i++)
229 if (first_width != s->streams[as->streams[i]]->codec->width)
235 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
237 static int check_matching_height(AVFormatContext *s, AdaptationSet *as) {
239 if (as->nb_streams < 2) return 1;
240 first_height = s->streams[as->streams[0]]->codec->height;
241 for (i = 1; i < as->nb_streams; i++)
242 if (first_height != s->streams[as->streams[i]]->codec->height)
248 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
250 static int check_matching_sample_rate(AVFormatContext *s, AdaptationSet *as) {
251 int first_sample_rate, i;
252 if (as->nb_streams < 2) return 1;
253 first_sample_rate = s->streams[as->streams[0]]->codec->sample_rate;
254 for (i = 1; i < as->nb_streams; i++)
255 if (first_sample_rate != s->streams[as->streams[i]]->codec->sample_rate)
260 static void free_adaptation_sets(AVFormatContext *s) {
261 WebMDashMuxContext *w = s->priv_data;
263 for (i = 0; i < w->nb_as; i++) {
264 av_freep(&w->as[i].streams);
271 * Parses a live header filename and computes the representation id,
272 * initialization pattern and the media pattern. Pass NULL if you don't want to
273 * compute any of those 3. Returns 0 on success and non-zero on failure.
275 * Name of the header file should conform to the following pattern:
276 * <file_description>_<representation_id>.hdr where <file_description> can be
277 * anything. The chunks should be named according to the following pattern:
278 * <file_description>_<representation_id>_<chunk_number>.chk
280 static int parse_filename(char *filename, char **representation_id,
281 char **initialization_pattern, char **media_pattern) {
282 char *underscore_pos = NULL;
283 char *period_pos = NULL;
284 char *temp_pos = NULL;
285 char *filename_str = av_strdup(filename);
286 if (!filename_str) return AVERROR(ENOMEM);
287 temp_pos = av_stristr(filename_str, "_");
289 underscore_pos = temp_pos + 1;
290 temp_pos = av_stristr(temp_pos + 1, "_");
292 if (!underscore_pos) return AVERROR_INVALIDDATA;
293 period_pos = av_stristr(underscore_pos, ".");
294 if (!period_pos) return AVERROR_INVALIDDATA;
295 *(underscore_pos - 1) = 0;
296 if (representation_id) {
297 *representation_id = av_malloc(period_pos - underscore_pos + 1);
298 if (!(*representation_id)) return AVERROR(ENOMEM);
299 av_strlcpy(*representation_id, underscore_pos, period_pos - underscore_pos + 1);
301 if (initialization_pattern) {
302 *initialization_pattern = av_asprintf("%s_$RepresentationID$.hdr",
304 if (!(*initialization_pattern)) return AVERROR(ENOMEM);
307 *media_pattern = av_asprintf("%s_$RepresentationID$_$Number$.chk",
309 if (!(*media_pattern)) return AVERROR(ENOMEM);
311 av_free(filename_str);
316 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
318 static int write_adaptation_set(AVFormatContext *s, int as_index)
320 WebMDashMuxContext *w = s->priv_data;
321 AdaptationSet *as = &w->as[as_index];
322 AVCodecContext *codec = s->streams[as->streams[0]]->codec;
323 AVDictionaryEntry *lang;
325 static const char boolean[2][6] = { "false", "true" };
326 int subsegmentStartsWithSAP = 1;
328 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
329 // are the same for all contained Representations. otherwise, they will go
330 // on their respective Representation tag. For live streams, they always go
331 // in the Representation tag.
332 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
333 if (codec->codec_type == AVMEDIA_TYPE_VIDEO) {
334 width_in_as = !w->is_live && check_matching_width(s, as);
335 height_in_as = !w->is_live && check_matching_height(s, as);
337 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
340 avio_printf(s->pb, "<AdaptationSet id=\"%s\"", as->id);
341 avio_printf(s->pb, " mimeType=\"%s/webm\"",
342 codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
343 avio_printf(s->pb, " codecs=\"%s\"", get_codec_name(codec->codec_id));
345 lang = av_dict_get(s->streams[as->streams[0]]->metadata, "language", NULL, 0);
346 if (lang) avio_printf(s->pb, " lang=\"%s\"", lang->value);
348 if (codec->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
349 avio_printf(s->pb, " width=\"%d\"", codec->width);
350 if (codec->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
351 avio_printf(s->pb, " height=\"%d\"", codec->height);
352 if (codec->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
353 avio_printf(s->pb, " audioSamplingRate=\"%d\"", codec->sample_rate);
355 avio_printf(s->pb, " bitstreamSwitching=\"%s\"",
356 boolean[bitstream_switching(s, as)]);
357 avio_printf(s->pb, " subsegmentAlignment=\"%s\"",
358 boolean[w->is_live || subsegment_alignment(s, as)]);
360 for (i = 0; i < as->nb_streams; i++) {
361 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
362 CLUSTER_KEYFRAME, NULL, 0);
363 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
365 avio_printf(s->pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
366 avio_printf(s->pb, ">\n");
369 AVDictionaryEntry *filename =
370 av_dict_get(s->streams[as->streams[0]]->metadata, FILENAME, NULL, 0);
371 char *initialization_pattern = NULL;
372 char *media_pattern = NULL;
373 int ret = parse_filename(filename->value, NULL, &initialization_pattern,
376 avio_printf(s->pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
377 codec->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
378 avio_printf(s->pb, "<SegmentTemplate");
379 avio_printf(s->pb, " timescale=\"1000\"");
380 avio_printf(s->pb, " duration=\"%d\"", w->chunk_duration);
381 avio_printf(s->pb, " media=\"%s\"", media_pattern);
382 avio_printf(s->pb, " startNumber=\"%d\"", w->chunk_start_index);
383 avio_printf(s->pb, " initialization=\"%s\"", initialization_pattern);
384 avio_printf(s->pb, "/>\n");
385 av_free(initialization_pattern);
386 av_free(media_pattern);
389 for (i = 0; i < as->nb_streams; i++) {
390 char *representation_id = NULL;
393 AVDictionaryEntry *filename =
394 av_dict_get(s->streams[as->streams[i]]->metadata, FILENAME, NULL, 0);
396 return AVERROR(EINVAL);
397 if (ret = parse_filename(filename->value, &representation_id, NULL, NULL))
400 representation_id = av_asprintf("%d", w->representation_id++);
401 if (!representation_id) return AVERROR(ENOMEM);
403 ret = write_representation(s, s->streams[as->streams[i]],
404 representation_id, !width_in_as,
405 !height_in_as, !sample_rate_in_as);
406 av_free(representation_id);
409 avio_printf(s->pb, "</AdaptationSet>\n");
413 static int to_integer(char *p, int len)
416 char *q = av_malloc(sizeof(char) * len);
418 return AVERROR(ENOMEM);
419 av_strlcpy(q, p, len);
425 static int parse_adaptation_sets(AVFormatContext *s)
427 WebMDashMuxContext *w = s->priv_data;
428 char *p = w->adaptation_sets;
430 enum { new_set, parsed_id, parsing_streams } state;
431 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
433 while (p < w->adaptation_sets + strlen(w->adaptation_sets)) {
436 else if (state == new_set && !strncmp(p, "id=", 3)) {
437 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
439 return AVERROR(ENOMEM);
442 w->as[w->nb_as - 1].nb_streams = 0;
443 w->as[w->nb_as - 1].streams = NULL;
444 p += 3; // consume "id="
445 q = w->as[w->nb_as - 1].id;
446 while (*p != ',') *q++ = *p++;
450 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
451 p += 8; // consume "streams="
452 state = parsing_streams;
453 } else if (state == parsing_streams) {
454 struct AdaptationSet *as = &w->as[w->nb_as - 1];
456 while (*q != '\0' && *q != ',' && *q != ' ') q++;
457 as->streams = av_realloc(as->streams, sizeof(*as->streams) * ++as->nb_streams);
458 if (as->streams == NULL)
459 return AVERROR(ENOMEM);
460 as->streams[as->nb_streams - 1] = to_integer(p, q - p + 1);
461 if (as->streams[as->nb_streams - 1] < 0) return -1;
462 if (*q == '\0') break;
463 if (*q == ' ') state = new_set;
472 static int webm_dash_manifest_write_header(AVFormatContext *s)
477 WebMDashMuxContext *w = s->priv_data;
478 ret = parse_adaptation_sets(s);
482 ret = write_header(s);
486 avio_printf(s->pb, "<Period id=\"0\"");
487 avio_printf(s->pb, " start=\"PT%gS\"", start);
489 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
491 avio_printf(s->pb, " >\n");
493 for (i = 0; i < w->nb_as; i++) {
494 ret = write_adaptation_set(s, i);
500 avio_printf(s->pb, "</Period>\n");
503 free_adaptation_sets(s);
504 return ret < 0 ? ret : 0;
507 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
512 static int webm_dash_manifest_write_trailer(AVFormatContext *s)
514 free_adaptation_sets(s);
518 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
519 static const AVOption options[] = {
520 { "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 },
521 { "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 },
522 { "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 },
523 { "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 },
524 { "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 },
525 { "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 },
526 { "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 },
527 { "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 },
531 #if CONFIG_WEBM_DASH_MANIFEST_MUXER
532 static const AVClass webm_dash_class = {
533 .class_name = "WebM DASH Manifest muxer",
534 .item_name = av_default_item_name,
536 .version = LIBAVUTIL_VERSION_INT,
539 AVOutputFormat ff_webm_dash_manifest_muxer = {
540 .name = "webm_dash_manifest",
541 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
542 .mime_type = "application/xml",
544 .priv_data_size = sizeof(WebMDashMuxContext),
545 .write_header = webm_dash_manifest_write_header,
546 .write_packet = webm_dash_manifest_write_packet,
547 .write_trailer = webm_dash_manifest_write_trailer,
548 .priv_class = &webm_dash_class,