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 AVIOContext *pb = s->pb;
83 double min_buffer_time = 1.0;
84 avio_printf(pb, "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n");
85 avio_printf(pb, "<MPD\n");
86 avio_printf(pb, " xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n");
87 avio_printf(pb, " xmlns=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
88 avio_printf(pb, " xsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011\"\n");
89 avio_printf(pb, " type=\"%s\"\n", w->is_live ? "dynamic" : "static");
91 avio_printf(pb, " mediaPresentationDuration=\"PT%gS\"\n",
94 avio_printf(pb, " minBufferTime=\"PT%gS\"\n", min_buffer_time);
95 avio_printf(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(pb, " availabilityStartTime=\"%s\"\n", gmt_iso);
110 avio_printf(pb, " timeShiftBufferDepth=\"PT%gS\"\n", w->time_shift_buffer_depth);
111 avio_printf(pb, " minimumUpdatePeriod=\"PT%dS\"", w->minimum_update_period);
112 avio_printf(pb, ">\n");
113 if (w->utc_timing_url) {
114 avio_printf(pb, "<UTCTiming\n");
115 avio_printf(pb, " schemeIdUri=\"urn:mpeg:dash:utc:http-iso:2014\"\n");
116 avio_printf(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, const AdaptationSet *as)
130 AVDictionaryEntry *gold = av_dict_get(s->streams[as->streams[0]]->metadata,
131 CUE_TIMESTAMPS, NULL, 0);
133 for (i = 1; i < as->nb_streams; i++) {
134 AVDictionaryEntry *ts = av_dict_get(s->streams[as->streams[i]]->metadata,
135 CUE_TIMESTAMPS, NULL, 0);
136 if (!ts || !av_strstart(ts->value, gold->value, NULL)) return 0;
141 static int bitstream_switching(AVFormatContext *s, const AdaptationSet *as)
144 const AVStream *gold_st = s->streams[as->streams[0]];
145 AVDictionaryEntry *gold_track_num = av_dict_get(gold_st->metadata,
146 TRACK_NUMBER, NULL, 0);
147 AVCodecParameters *gold_par = gold_st->codecpar;
148 if (!gold_track_num) return 0;
149 for (i = 1; i < as->nb_streams; i++) {
150 const AVStream *st = s->streams[as->streams[i]];
151 AVDictionaryEntry *track_num = av_dict_get(st->metadata,
152 TRACK_NUMBER, NULL, 0);
153 AVCodecParameters *par = st->codecpar;
155 !av_strstart(track_num->value, gold_track_num->value, NULL) ||
156 gold_par->codec_id != par->codec_id ||
157 gold_par->extradata_size != par->extradata_size ||
158 memcmp(gold_par->extradata, par->extradata, par->extradata_size)) {
166 * Writes a Representation within an Adaptation Set. Returns 0 on success and
169 static int write_representation(AVFormatContext *s, AVStream *st, char *id,
170 int output_width, int output_height,
171 int output_sample_rate)
173 WebMDashMuxContext *w = s->priv_data;
174 AVIOContext *pb = s->pb;
175 const AVCodecParameters *par = st->codecpar;
176 AVDictionaryEntry *bandwidth = av_dict_get(st->metadata, BANDWIDTH, NULL, 0);
177 const char *bandwidth_str;
178 avio_printf(pb, "<Representation id=\"%s\"", id);
180 bandwidth_str = bandwidth->value;
181 } else if (w->is_live) {
182 // if bandwidth for live was not provided, use a default
183 bandwidth_str = (par->codec_type == AVMEDIA_TYPE_AUDIO) ? "128000" : "1000000";
185 return AVERROR(EINVAL);
187 avio_printf(pb, " bandwidth=\"%s\"", bandwidth_str);
188 if (par->codec_type == AVMEDIA_TYPE_VIDEO && output_width)
189 avio_printf(pb, " width=\"%d\"", par->width);
190 if (par->codec_type == AVMEDIA_TYPE_VIDEO && output_height)
191 avio_printf(pb, " height=\"%d\"", par->height);
192 if (par->codec_type == AVMEDIA_TYPE_AUDIO && output_sample_rate)
193 avio_printf(pb, " audioSamplingRate=\"%d\"", par->sample_rate);
195 // For live streams, Codec and Mime Type always go in the Representation tag.
196 avio_printf(pb, " codecs=\"%s\"", get_codec_name(par->codec_id));
197 avio_printf(pb, " mimeType=\"%s/webm\"",
198 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
199 // For live streams, subsegments always start with key frames. So this
201 avio_printf(pb, " startsWithSAP=\"1\"");
202 avio_printf(pb, ">");
204 AVDictionaryEntry *irange = av_dict_get(st->metadata, INITIALIZATION_RANGE, NULL, 0);
205 AVDictionaryEntry *cues_start = av_dict_get(st->metadata, CUES_START, NULL, 0);
206 AVDictionaryEntry *cues_end = av_dict_get(st->metadata, CUES_END, NULL, 0);
207 AVDictionaryEntry *filename = av_dict_get(st->metadata, FILENAME, NULL, 0);
208 if (!irange || !cues_start || !cues_end || !filename)
209 return AVERROR(EINVAL);
211 avio_printf(pb, ">\n");
212 avio_printf(pb, "<BaseURL>%s</BaseURL>\n", filename->value);
213 avio_printf(pb, "<SegmentBase\n");
214 avio_printf(pb, " indexRange=\"%s-%s\">\n", cues_start->value, cues_end->value);
215 avio_printf(pb, "<Initialization\n");
216 avio_printf(pb, " range=\"0-%s\" />\n", irange->value);
217 avio_printf(pb, "</SegmentBase>\n");
219 avio_printf(pb, "</Representation>\n");
224 * Checks if width of all streams are the same. Returns 1 if true, 0 otherwise.
226 static int check_matching_width(AVFormatContext *s, const AdaptationSet *as)
229 if (as->nb_streams < 2) return 1;
230 first_width = s->streams[as->streams[0]]->codecpar->width;
231 for (i = 1; i < as->nb_streams; i++)
232 if (first_width != s->streams[as->streams[i]]->codecpar->width)
238 * Checks if height of all streams are the same. Returns 1 if true, 0 otherwise.
240 static int check_matching_height(AVFormatContext *s, const AdaptationSet *as)
243 if (as->nb_streams < 2) return 1;
244 first_height = s->streams[as->streams[0]]->codecpar->height;
245 for (i = 1; i < as->nb_streams; i++)
246 if (first_height != s->streams[as->streams[i]]->codecpar->height)
252 * Checks if sample rate of all streams are the same. Returns 1 if true, 0 otherwise.
254 static int check_matching_sample_rate(AVFormatContext *s, const 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)
267 WebMDashMuxContext *w = s->priv_data;
269 for (i = 0; i < w->nb_as; i++) {
270 av_freep(&w->as[i].streams);
277 * Parses a live header filename and returns the position of the '_' and '.'
278 * delimiting <file_description> and <representation_id>.
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 split_filename(char *filename, char **underscore_pos,
288 *underscore_pos = strrchr(filename, '_');
289 if (!*underscore_pos)
290 return AVERROR(EINVAL);
291 *period_pos = strchr(*underscore_pos, '.');
293 return AVERROR(EINVAL);
298 * Writes an Adaptation Set. Returns 0 on success and < 0 on failure.
300 static int write_adaptation_set(AVFormatContext *s, int as_index)
302 WebMDashMuxContext *w = s->priv_data;
303 AdaptationSet *as = &w->as[as_index];
304 const AVStream *st = s->streams[as->streams[0]];
305 AVCodecParameters *par = st->codecpar;
306 AVDictionaryEntry *lang;
307 AVIOContext *pb = s->pb;
309 static const char boolean[2][6] = { "false", "true" };
310 int subsegmentStartsWithSAP = 1;
312 // Width, Height and Sample Rate will go in the AdaptationSet tag if they
313 // are the same for all contained Representations. otherwise, they will go
314 // on their respective Representation tag. For live streams, they always go
315 // in the Representation tag.
316 int width_in_as = 1, height_in_as = 1, sample_rate_in_as = 1;
317 if (par->codec_type == AVMEDIA_TYPE_VIDEO) {
318 width_in_as = !w->is_live && check_matching_width (s, as);
319 height_in_as = !w->is_live && check_matching_height(s, as);
321 sample_rate_in_as = !w->is_live && check_matching_sample_rate(s, as);
324 avio_printf(pb, "<AdaptationSet id=\"%s\"", as->id);
325 avio_printf(pb, " mimeType=\"%s/webm\"",
326 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
327 avio_printf(pb, " codecs=\"%s\"", get_codec_name(par->codec_id));
329 lang = av_dict_get(st->metadata, "language", NULL, 0);
331 avio_printf(pb, " lang=\"%s\"", lang->value);
333 if (par->codec_type == AVMEDIA_TYPE_VIDEO && width_in_as)
334 avio_printf(pb, " width=\"%d\"", par->width);
335 if (par->codec_type == AVMEDIA_TYPE_VIDEO && height_in_as)
336 avio_printf(pb, " height=\"%d\"", par->height);
337 if (par->codec_type == AVMEDIA_TYPE_AUDIO && sample_rate_in_as)
338 avio_printf(pb, " audioSamplingRate=\"%d\"", par->sample_rate);
340 avio_printf(pb, " bitstreamSwitching=\"%s\"",
341 boolean[bitstream_switching(s, as)]);
342 avio_printf(pb, " subsegmentAlignment=\"%s\"",
343 boolean[w->is_live || subsegment_alignment(s, as)]);
345 for (i = 0; i < as->nb_streams; i++) {
346 AVDictionaryEntry *kf = av_dict_get(s->streams[as->streams[i]]->metadata,
347 CLUSTER_KEYFRAME, NULL, 0);
348 if (!w->is_live && (!kf || !strncmp(kf->value, "0", 1))) subsegmentStartsWithSAP = 0;
350 avio_printf(pb, " subsegmentStartsWithSAP=\"%d\"", subsegmentStartsWithSAP);
351 avio_printf(pb, ">\n");
354 AVDictionaryEntry *filename =
355 av_dict_get(st->metadata, FILENAME, NULL, 0);
356 char *underscore_pos, *period_pos;
359 return AVERROR(EINVAL);
360 ret = split_filename(filename->value, &underscore_pos, &period_pos);
362 *underscore_pos = '\0';
363 avio_printf(pb, "<ContentComponent id=\"1\" type=\"%s\"/>\n",
364 par->codec_type == AVMEDIA_TYPE_VIDEO ? "video" : "audio");
365 avio_printf(pb, "<SegmentTemplate");
366 avio_printf(pb, " timescale=\"1000\"");
367 avio_printf(pb, " duration=\"%d\"", w->chunk_duration);
368 avio_printf(pb, " media=\"%s_$RepresentationID$_$Number$.chk\"",
370 avio_printf(pb, " startNumber=\"%d\"", w->chunk_start_index);
371 avio_printf(pb, " initialization=\"%s_$RepresentationID$.hdr\"",
373 avio_printf(pb, "/>\n");
374 *underscore_pos = '_';
377 for (i = 0; i < as->nb_streams; i++) {
378 char buf[25], *representation_id = buf, *underscore_pos, *period_pos;
379 AVStream *st = s->streams[as->streams[i]];
382 AVDictionaryEntry *filename =
383 av_dict_get(st->metadata, FILENAME, NULL, 0);
385 return AVERROR(EINVAL);
386 ret = split_filename(filename->value, &underscore_pos, &period_pos);
389 representation_id = underscore_pos + 1;
392 snprintf(buf, sizeof(buf), "%d", w->representation_id++);
394 ret = write_representation(s, st, representation_id, !width_in_as,
395 !height_in_as, !sample_rate_in_as);
400 avio_printf(s->pb, "</AdaptationSet>\n");
404 static int parse_adaptation_sets(AVFormatContext *s)
406 WebMDashMuxContext *w = s->priv_data;
407 char *p = w->adaptation_sets;
409 enum { new_set, parsed_id, parsing_streams } state;
410 if (!w->adaptation_sets) {
411 av_log(s, AV_LOG_ERROR, "The 'adaptation_sets' option must be set.\n");
412 return AVERROR(EINVAL);
414 // syntax id=0,streams=0,1,2 id=1,streams=3,4 and so on
418 if (state == new_set)
421 return AVERROR(EINVAL);
422 } else if (state == new_set && *p == ' ') {
425 } else if (state == new_set && !strncmp(p, "id=", 3)) {
426 void *mem = av_realloc(w->as, sizeof(*w->as) * (w->nb_as + 1));
429 return AVERROR(ENOMEM);
432 w->as[w->nb_as - 1].nb_streams = 0;
433 w->as[w->nb_as - 1].streams = NULL;
434 p += 3; // consume "id="
435 q = w->as[w->nb_as - 1].id;
436 comma = strchr(p, ',');
437 if (!comma || comma - p >= sizeof(w->as[w->nb_as - 1].id)) {
438 av_log(s, AV_LOG_ERROR, "'id' in 'adaptation_sets' is malformed.\n");
439 return AVERROR(EINVAL);
441 while (*p != ',') *q++ = *p++;
445 } else if (state == parsed_id && !strncmp(p, "streams=", 8)) {
446 p += 8; // consume "streams="
447 state = parsing_streams;
448 } else if (state == parsing_streams) {
449 struct AdaptationSet *as = &w->as[w->nb_as - 1];
451 int ret = av_reallocp_array(&as->streams, ++as->nb_streams,
452 sizeof(*as->streams));
455 num = strtoll(p, &q, 10);
456 if (!av_isdigit(*p) || (*q != ' ' && *q != '\0' && *q != ',') ||
457 num < 0 || num >= s->nb_streams) {
458 av_log(s, AV_LOG_ERROR, "Invalid value for 'streams' in adapation_sets.\n");
459 return AVERROR(EINVAL);
461 as->streams[as->nb_streams - 1] = num;
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;
479 for (unsigned i = 0; i < s->nb_streams; i++) {
480 enum AVCodecID codec_id = s->streams[i]->codecpar->codec_id;
481 if (codec_id != AV_CODEC_ID_VP8 && codec_id != AV_CODEC_ID_VP9 &&
482 codec_id != AV_CODEC_ID_VORBIS && codec_id != AV_CODEC_ID_OPUS)
483 return AVERROR(EINVAL);
486 ret = parse_adaptation_sets(s);
490 ret = write_header(s);
494 avio_printf(s->pb, "<Period id=\"0\"");
495 avio_printf(s->pb, " start=\"PT%gS\"", start);
497 avio_printf(s->pb, " duration=\"PT%gS\"", get_duration(s));
499 avio_printf(s->pb, " >\n");
501 for (i = 0; i < w->nb_as; i++) {
502 ret = write_adaptation_set(s, i);
508 avio_printf(s->pb, "</Period>\n");
511 free_adaptation_sets(s);
512 return ret < 0 ? ret : 0;
515 static int webm_dash_manifest_write_packet(AVFormatContext *s, AVPacket *pkt)
520 #define OFFSET(x) offsetof(WebMDashMuxContext, x)
521 static const AVOption options[] = {
522 { "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 },
523 { "live", "create a live stream manifest", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM },
524 { "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 },
525 { "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 },
526 { "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 },
527 { "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 },
528 { "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 },
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 .priv_class = &webm_dash_class,