2 * Copyright (c) 2012 Martin Storsjo
4 * This file is part of Libav.
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * To create a simple file for smooth streaming:
23 * avconv <normal input/transcoding options> -movflags frag_keyframe foo.ismv
24 * ismindex -n foo foo.ismv
25 * This step creates foo.ism and foo.ismc that is required by IIS for
28 * To pre-split files for serving as static files by a web server without
29 * any extra server support, create the ismv file as above, and split it:
30 * ismindex -split foo.ismv
31 * This step creates a file Manifest and directories QualityLevel(...),
32 * that can be read directly by a smooth streaming player.
40 #define mkdir(a, b) _mkdir(a)
43 #include "libavformat/avformat.h"
44 #include "libavutil/intreadwrite.h"
45 #include "libavutil/mathematics.h"
47 static int usage(const char *argv0, int ret)
49 fprintf(stderr, "%s [-split] [-n basename] file1 [file2] ...\n", argv0);
64 int is_audio, is_video;
67 int sample_rate, channels;
68 uint8_t *codec_private;
69 int codec_private_size;
70 struct MoofOffset *offsets;
80 struct VideoFile **files;
81 int video_file, audio_file;
82 int nb_video_files, nb_audio_files;
85 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
98 int len = FFMIN(sizeof(buf), size);
99 if (avio_read(in, buf, len) != len)
101 avio_write(out, buf, len);
107 static int write_fragment(const char *filename, AVIOContext *in)
109 AVIOContext *out = NULL;
112 if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
114 copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
115 copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
123 static int write_fragments(struct VideoFiles *files, int start_index,
126 char dirname[100], filename[500];
129 for (i = start_index; i < files->nb_files; i++) {
130 struct VideoFile *vf = files->files[i];
131 const char *type = vf->is_video ? "video" : "audio";
132 snprintf(dirname, sizeof(dirname), "QualityLevels(%d)", vf->bitrate);
133 mkdir(dirname, 0777);
134 for (j = 0; j < vf->chunks; j++) {
135 snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
136 dirname, type, vf->offsets[j].time);
137 avio_seek(in, vf->offsets[j].offset, SEEK_SET);
138 write_fragment(filename, in);
144 static int read_tfra(struct VideoFiles *files, int start_index, AVIOContext *f)
146 int ret = AVERROR_EOF, track_id;
147 int version, fieldlength, i, j;
148 int64_t pos = avio_tell(f);
149 uint32_t size = avio_rb32(f);
150 struct VideoFile *vf = NULL;
152 if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
154 version = avio_r8(f);
156 track_id = avio_rb32(f); /* track id */
157 for (i = start_index; i < files->nb_files && !vf; i++)
158 if (files->files[i]->track_id == track_id)
159 vf = files->files[i];
161 /* Ok, continue parsing the next atom */
165 fieldlength = avio_rb32(f);
166 vf->chunks = avio_rb32(f);
167 vf->offsets = av_mallocz(sizeof(*vf->offsets) * vf->chunks);
169 ret = AVERROR(ENOMEM);
172 for (i = 0; i < vf->chunks; i++) {
174 vf->offsets[i].time = avio_rb64(f);
175 vf->offsets[i].offset = avio_rb64(f);
177 vf->offsets[i].time = avio_rb32(f);
178 vf->offsets[i].offset = avio_rb32(f);
180 for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
182 for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
184 for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
187 vf->offsets[i - 1].duration = vf->offsets[i].time -
188 vf->offsets[i - 1].time;
191 vf->offsets[vf->chunks - 1].duration = vf->duration -
192 vf->offsets[vf->chunks - 1].time;
196 avio_seek(f, pos + size, SEEK_SET);
200 static int read_mfra(struct VideoFiles *files, int start_index,
201 const char *file, int split)
204 AVIOContext *f = NULL;
207 if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
209 avio_seek(f, avio_size(f) - 4, SEEK_SET);
210 mfra_size = avio_rb32(f);
211 avio_seek(f, -mfra_size, SEEK_CUR);
212 if (avio_rb32(f) != mfra_size) {
213 err = AVERROR_INVALIDDATA;
216 if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
217 err = AVERROR_INVALIDDATA;
220 while (!read_tfra(files, start_index, f)) {
225 write_fragments(files, start_index, f);
231 fprintf(stderr, "Unable to read the MFRA atom in %s\n", file);
235 static int get_private_data(struct VideoFile *vf, AVCodecContext *codec)
237 vf->codec_private_size = codec->extradata_size;
238 vf->codec_private = av_mallocz(codec->extradata_size);
239 if (!vf->codec_private)
240 return AVERROR(ENOMEM);
241 memcpy(vf->codec_private, codec->extradata, codec->extradata_size);
245 static int get_video_private_data(struct VideoFile *vf, AVCodecContext *codec)
247 AVIOContext *io = NULL;
248 uint16_t sps_size, pps_size;
249 int err = AVERROR(EINVAL);
251 if (codec->codec_id == AV_CODEC_ID_VC1)
252 return get_private_data(vf, codec);
254 avio_open_dyn_buf(&io);
255 if (codec->extradata_size < 11 || codec->extradata[0] != 1)
257 sps_size = AV_RB16(&codec->extradata[6]);
258 if (11 + sps_size > codec->extradata_size)
260 avio_wb32(io, 0x00000001);
261 avio_write(io, &codec->extradata[8], sps_size);
262 pps_size = AV_RB16(&codec->extradata[9 + sps_size]);
263 if (11 + sps_size + pps_size > codec->extradata_size)
265 avio_wb32(io, 0x00000001);
266 avio_write(io, &codec->extradata[11 + sps_size], pps_size);
270 vf->codec_private_size = avio_close_dyn_buf(io, &vf->codec_private);
274 static int handle_file(struct VideoFiles *files, const char *file, int split)
276 AVFormatContext *ctx = NULL;
277 int err = 0, i, orig_files = files->nb_files;
278 char errbuf[50], *ptr;
279 struct VideoFile *vf;
281 err = avformat_open_input(&ctx, file, NULL, NULL);
283 av_strerror(err, errbuf, sizeof(errbuf));
284 fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
288 err = avformat_find_stream_info(ctx, NULL);
290 av_strerror(err, errbuf, sizeof(errbuf));
291 fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
295 if (ctx->nb_streams < 1) {
296 fprintf(stderr, "No streams found in %s\n", file);
299 if (!files->duration)
300 files->duration = ctx->duration;
302 for (i = 0; i < ctx->nb_streams; i++) {
303 AVStream *st = ctx->streams[i];
304 vf = av_mallocz(sizeof(*vf));
305 files->files = av_realloc(files->files,
306 sizeof(*files->files) * (files->nb_files + 1));
307 files->files[files->nb_files] = vf;
310 if ((ptr = strrchr(file, '/')) != NULL)
313 vf->bitrate = st->codec->bit_rate;
314 vf->track_id = st->id;
315 vf->timescale = st->time_base.den;
316 vf->duration = av_rescale_rnd(ctx->duration, vf->timescale,
317 AV_TIME_BASE, AV_ROUND_UP);
318 vf->is_audio = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
319 vf->is_video = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
321 if (!vf->is_audio && !vf->is_video) {
323 "Track %d in %s is neither video nor audio, skipping\n",
325 av_freep(&files->files[files->nb_files]);
330 if (files->audio_file < 0)
331 files->audio_file = files->nb_files;
332 files->nb_audio_files++;
333 vf->channels = st->codec->channels;
334 vf->sample_rate = st->codec->sample_rate;
335 if (st->codec->codec_id == AV_CODEC_ID_AAC) {
339 } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
341 vf->tag = st->codec->codec_tag;
342 vf->blocksize = st->codec->block_align;
344 get_private_data(vf, st->codec);
347 if (files->video_file < 0)
348 files->video_file = files->nb_files;
349 files->nb_video_files++;
350 vf->width = st->codec->width;
351 vf->height = st->codec->height;
352 if (st->codec->codec_id == AV_CODEC_ID_H264)
354 else if (st->codec->codec_id == AV_CODEC_ID_VC1)
356 get_video_private_data(vf, st->codec);
362 avformat_close_input(&ctx);
364 err = read_mfra(files, orig_files, file, split);
368 avformat_close_input(&ctx);
372 static void output_server_manifest(struct VideoFiles *files,
373 const char *basename)
379 snprintf(filename, sizeof(filename), "%s.ism", basename);
380 out = fopen(filename, "w");
385 fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
386 fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
387 fprintf(out, "\t<head>\n");
388 fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
389 "content=\"%s.ismc\" />\n", basename);
390 fprintf(out, "\t</head>\n");
391 fprintf(out, "\t<body>\n");
392 fprintf(out, "\t\t<switch>\n");
393 for (i = 0; i < files->nb_files; i++) {
394 struct VideoFile *vf = files->files[i];
395 const char *type = vf->is_video ? "video" : "audio";
396 fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
397 type, vf->name, vf->bitrate);
398 fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
399 "valueType=\"data\" />\n", vf->track_id);
400 fprintf(out, "\t\t\t</%s>\n", type);
402 fprintf(out, "\t\t</switch>\n");
403 fprintf(out, "\t</body>\n");
404 fprintf(out, "</smil>\n");
408 static void output_client_manifest(struct VideoFiles *files,
409 const char *basename, int split)
416 snprintf(filename, sizeof(filename), "Manifest");
418 snprintf(filename, sizeof(filename), "%s.ismc", basename);
419 out = fopen(filename, "w");
424 fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
425 fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
426 "Duration=\"%"PRId64 "\">\n", files->duration * 10);
427 if (files->video_file >= 0) {
428 struct VideoFile *vf = files->files[files->video_file];
429 struct VideoFile *first_vf = vf;
432 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
434 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
435 files->nb_video_files, vf->chunks);
436 for (i = 0; i < files->nb_files; i++) {
437 vf = files->files[i];
441 "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
442 "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
443 "CodecPrivateData=\"",
444 index, vf->bitrate, vf->fourcc, vf->width, vf->height);
445 for (j = 0; j < vf->codec_private_size; j++)
446 fprintf(out, "%02X", vf->codec_private[j]);
447 fprintf(out, "\" />\n");
449 if (vf->chunks != first_vf->chunks)
450 fprintf(stderr, "Mismatched number of video chunks in %s and %s\n",
451 vf->name, first_vf->name);
454 for (i = 0; i < vf->chunks; i++) {
455 for (j = files->video_file + 1; j < files->nb_files; j++) {
456 if (files->files[j]->is_video &&
457 vf->offsets[i].duration != files->files[j]->offsets[i].duration)
458 fprintf(stderr, "Mismatched duration of video chunk %d in %s and %s\n",
459 i, vf->name, files->files[j]->name);
461 fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n", i,
462 vf->offsets[i].duration);
464 fprintf(out, "\t</StreamIndex>\n");
466 if (files->audio_file >= 0) {
467 struct VideoFile *vf = files->files[files->audio_file];
468 struct VideoFile *first_vf = vf;
471 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
473 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
474 files->nb_audio_files, vf->chunks);
475 for (i = 0; i < files->nb_files; i++) {
476 vf = files->files[i];
480 "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
481 "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
482 "BitsPerSample=\"16\" PacketSize=\"%d\" "
483 "AudioTag=\"%d\" CodecPrivateData=\"",
484 index, vf->bitrate, vf->fourcc, vf->sample_rate,
485 vf->channels, vf->blocksize, vf->tag);
486 for (j = 0; j < vf->codec_private_size; j++)
487 fprintf(out, "%02X", vf->codec_private[j]);
488 fprintf(out, "\" />\n");
490 if (vf->chunks != first_vf->chunks)
491 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
492 vf->name, first_vf->name);
495 for (i = 0; i < vf->chunks; i++) {
496 for (j = files->audio_file + 1; j < files->nb_files; j++) {
497 if (files->files[j]->is_audio &&
498 vf->offsets[i].duration != files->files[j]->offsets[i].duration)
499 fprintf(stderr, "Mismatched duration of audio chunk %d in %s and %s\n",
500 i, vf->name, files->files[j]->name);
502 fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
503 i, vf->offsets[i].duration);
505 fprintf(out, "\t</StreamIndex>\n");
507 fprintf(out, "</SmoothStreamingMedia>\n");
511 static void clean_files(struct VideoFiles *files)
514 for (i = 0; i < files->nb_files; i++) {
515 av_freep(&files->files[i]->codec_private);
516 av_freep(&files->files[i]->offsets);
517 av_freep(&files->files[i]);
519 av_freep(&files->files);
523 int main(int argc, char **argv)
525 const char *basename = NULL;
527 struct VideoFiles vf = { 0, .video_file = -1, .audio_file = -1 };
531 for (i = 1; i < argc; i++) {
532 if (!strcmp(argv[i], "-n")) {
533 basename = argv[i + 1];
535 } else if (!strcmp(argv[i], "-split")) {
537 } else if (argv[i][0] == '-') {
538 return usage(argv[0], 1);
540 if (handle_file(&vf, argv[i], split))
544 if (!vf.nb_files || (!basename && !split))
545 return usage(argv[0], 1);
548 output_server_manifest(&vf, basename);
549 output_client_manifest(&vf, basename, split);