]> git.sesse.net Git - ffmpeg/blobdiff - tools/ismindex.c
screenpresso: Add extended pixel format support
[ffmpeg] / tools / ismindex.c
index 5980869c24dc8f4dc19e66d14115f69a058005e4..71bb36bf530c7d92ba2815a15bb4b6af5c227b0b 100644 (file)
  * This step creates foo.ism and foo.ismc that is required by IIS for
  * serving it.
  *
+ * With -ismf, it also creates foo.ismf, which maps fragment names to
+ * start-end offsets in the ismv, for use in your own streaming server.
+ *
+ * By adding -path-prefix path/, the produced foo.ism will refer to the
+ * files foo.ismv as "path/foo.ismv" - the prefix for the generated ismc
+ * file can be set with the -ismc-prefix option similarly.
+ *
  * To pre-split files for serving as static files by a web server without
  * any extra server support, create the ismv file as above, and split it:
  * ismindex -split foo.ismv
  * This step creates a file Manifest and directories QualityLevel(...),
  * that can be read directly by a smooth streaming player.
+ *
+ * The -output dir option can be used to request that output files
+ * (both .ism/.ismc, or Manifest/QualityLevels* when splitting)
+ * should be written to this directory instead of in the current directory.
+ * (The directory itself isn't created if it doesn't already exist.)
  */
 
 #include <stdio.h>
 #include <string.h>
-#include <sys/stat.h>
-#ifdef _WIN32
-#include <io.h>
-#define mkdir(a, b) mkdir(a)
-#endif
 
 #include "libavformat/avformat.h"
+#include "libavformat/isom.h"
+#include "libavformat/os_support.h"
 #include "libavutil/intreadwrite.h"
 #include "libavutil/mathematics.h"
 
 static int usage(const char *argv0, int ret)
 {
-    fprintf(stderr, "%s [-split] [-n basename] file1 [file2] ...\n", argv0);
+    fprintf(stderr, "%s [-split] [-ismf] [-n basename] [-path-prefix prefix] "
+                    "[-ismc-prefix prefix] [-output dir] file1 [file2] ...\n", argv0);
     return ret;
 }
 
 struct MoofOffset {
     int64_t time;
     int64_t offset;
-    int duration;
+    int64_t duration;
 };
 
-struct VideoFile {
+struct Track {
     const char *name;
     int64_t duration;
     int bitrate;
@@ -74,14 +84,26 @@ struct VideoFile {
     int tag;
 };
 
-struct VideoFiles {
-    int nb_files;
+struct Tracks {
+    int nb_tracks;
     int64_t duration;
-    struct VideoFile **files;
-    int video_file, audio_file;
-    int nb_video_files, nb_audio_files;
+    struct Track **tracks;
+    int video_track, audio_track;
+    int nb_video_tracks, nb_audio_tracks;
 };
 
+static int expect_tag(int32_t got_tag, int32_t expected_tag) {
+    if (got_tag != expected_tag) {
+        char got_tag_str[4], expected_tag_str[4];
+        AV_WB32(got_tag_str, got_tag);
+        AV_WB32(expected_tag_str, expected_tag);
+        fprintf(stderr, "wanted tag %.4s, got %.4s\n", expected_tag_str,
+                got_tag_str);
+        return -1;
+    }
+    return 0;
+}
+
 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
 {
     int32_t size, tag;
@@ -90,29 +112,50 @@ static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
     tag  = avio_rb32(in);
     avio_wb32(out, size);
     avio_wb32(out, tag);
-    if (tag != tag_name)
+    if (expect_tag(tag, tag_name) != 0)
         return -1;
     size -= 8;
     while (size > 0) {
         char buf[1024];
         int len = FFMIN(sizeof(buf), size);
-        if (avio_read(in, buf, len) != len)
+        int got;
+        if ((got = avio_read(in, buf, len)) != len) {
+            fprintf(stderr, "short read, wanted %d, got %d\n", len, got);
             break;
+        }
         avio_write(out, buf, len);
         size -= len;
     }
     return 0;
 }
 
+static int skip_tag(AVIOContext *in, int32_t tag_name)
+{
+    int64_t pos = avio_tell(in);
+    int32_t size, tag;
+
+    size = avio_rb32(in);
+    tag  = avio_rb32(in);
+    if (expect_tag(tag, tag_name) != 0)
+        return -1;
+    avio_seek(in, pos + size, SEEK_SET);
+    return 0;
+}
+
 static int write_fragment(const char *filename, AVIOContext *in)
 {
     AVIOContext *out = NULL;
     int ret;
 
-    if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
+    if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0) {
+        char errbuf[100];
+        av_strerror(ret, errbuf, sizeof(errbuf));
+        fprintf(stderr, "Unable to open %s: %s\n", filename, errbuf);
         return ret;
-    copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
-    copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
+    }
+    ret = copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
+    if (!ret)
+        ret = copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
 
     avio_flush(out);
     avio_close(out);
@@ -120,62 +163,199 @@ static int write_fragment(const char *filename, AVIOContext *in)
     return ret;
 }
 
-static int write_fragments(struct VideoFiles *files, int start_index,
-                           AVIOContext *in)
+static int skip_fragment(AVIOContext *in)
 {
-    char dirname[100], filename[500];
-    int i, j;
+    int ret;
+    ret = skip_tag(in, MKBETAG('m', 'o', 'o', 'f'));
+    if (!ret)
+        ret = skip_tag(in, MKBETAG('m', 'd', 'a', 't'));
+    return ret;
+}
 
-    for (i = start_index; i < files->nb_files; i++) {
-        struct VideoFile *vf = files->files[i];
-        const char *type     = vf->is_video ? "video" : "audio";
-        snprintf(dirname, sizeof(dirname), "QualityLevels(%d)", vf->bitrate);
-        mkdir(dirname, 0777);
-        for (j = 0; j < vf->chunks; j++) {
+static int write_fragments(struct Tracks *tracks, int start_index,
+                           AVIOContext *in, const char *basename,
+                           int split, int ismf, const char* output_prefix)
+{
+    char dirname[2048], filename[2048], idxname[2048];
+    int i, j, ret = 0, fragment_ret;
+    FILE* out = NULL;
+
+    if (ismf) {
+        snprintf(idxname, sizeof(idxname), "%s%s.ismf", output_prefix, basename);
+        out = fopen(idxname, "w");
+        if (!out) {
+            ret = AVERROR(errno);
+            perror(idxname);
+            goto fail;
+        }
+    }
+    for (i = start_index; i < tracks->nb_tracks; i++) {
+        struct Track *track = tracks->tracks[i];
+        const char *type    = track->is_video ? "video" : "audio";
+        snprintf(dirname, sizeof(dirname), "%sQualityLevels(%d)", output_prefix, track->bitrate);
+        if (split) {
+            if (mkdir(dirname, 0777) == -1 && errno != EEXIST) {
+                ret = AVERROR(errno);
+                perror(dirname);
+                goto fail;
+            }
+        }
+        for (j = 0; j < track->chunks; j++) {
             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
-                     dirname, type, vf->offsets[j].time);
-            avio_seek(in, vf->offsets[j].offset, SEEK_SET);
-            write_fragment(filename, in);
+                     dirname, type, track->offsets[j].time);
+            avio_seek(in, track->offsets[j].offset, SEEK_SET);
+            if (ismf)
+                fprintf(out, "%s %"PRId64, filename, avio_tell(in));
+            if (split)
+                fragment_ret = write_fragment(filename, in);
+            else
+                fragment_ret = skip_fragment(in);
+            if (ismf)
+                fprintf(out, " %"PRId64"\n", avio_tell(in));
+            if (fragment_ret != 0) {
+                fprintf(stderr, "failed fragment %d in track %d (%s)\n", j,
+                        track->track_id, track->name);
+                ret = fragment_ret;
+            }
         }
     }
-    return 0;
+fail:
+    if (out)
+        fclose(out);
+    return ret;
+}
+
+static int64_t read_trun_duration(AVIOContext *in, int default_duration,
+                                  int64_t end)
+{
+    int64_t dts = 0;
+    int64_t pos;
+    int flags, i;
+    int entries;
+    int64_t first_pts = 0;
+    int64_t max_pts = 0;
+    avio_r8(in); /* version */
+    flags = avio_rb24(in);
+    if (default_duration <= 0 && !(flags & MOV_TRUN_SAMPLE_DURATION)) {
+        fprintf(stderr, "No sample duration in trun flags\n");
+        return -1;
+    }
+    entries = avio_rb32(in);
+
+    if (flags & MOV_TRUN_DATA_OFFSET)        avio_rb32(in);
+    if (flags & MOV_TRUN_FIRST_SAMPLE_FLAGS) avio_rb32(in);
+
+    pos = avio_tell(in);
+    for (i = 0; i < entries && pos < end; i++) {
+        int sample_duration = default_duration;
+        int64_t pts = dts;
+        if (flags & MOV_TRUN_SAMPLE_DURATION) sample_duration = avio_rb32(in);
+        if (flags & MOV_TRUN_SAMPLE_SIZE)     avio_rb32(in);
+        if (flags & MOV_TRUN_SAMPLE_FLAGS)    avio_rb32(in);
+        if (flags & MOV_TRUN_SAMPLE_CTS)      pts += avio_rb32(in);
+        if (sample_duration < 0) {
+            fprintf(stderr, "Negative sample duration %d\n", sample_duration);
+            return -1;
+        }
+        if (i == 0)
+            first_pts = pts;
+        max_pts = FFMAX(max_pts, pts + sample_duration);
+        dts += sample_duration;
+        pos = avio_tell(in);
+    }
+
+    return max_pts - first_pts;
+}
+
+static int64_t read_moof_duration(AVIOContext *in, int64_t offset)
+{
+    int64_t ret = -1;
+    int32_t moof_size, size, tag;
+    int64_t pos = 0;
+    int default_duration = 0;
+
+    avio_seek(in, offset, SEEK_SET);
+    moof_size = avio_rb32(in);
+    tag  = avio_rb32(in);
+    if (expect_tag(tag, MKBETAG('m', 'o', 'o', 'f')) != 0)
+        goto fail;
+    while (pos < offset + moof_size) {
+        pos = avio_tell(in);
+        size = avio_rb32(in);
+        tag  = avio_rb32(in);
+        if (tag == MKBETAG('t', 'r', 'a', 'f')) {
+            int64_t traf_pos = pos;
+            int64_t traf_size = size;
+            while (pos < traf_pos + traf_size) {
+                pos = avio_tell(in);
+                size = avio_rb32(in);
+                tag  = avio_rb32(in);
+                if (tag == MKBETAG('t', 'f', 'h', 'd')) {
+                    int flags = 0;
+                    avio_r8(in); /* version */
+                    flags = avio_rb24(in);
+                    avio_rb32(in); /* track_id */
+                    if (flags & MOV_TFHD_BASE_DATA_OFFSET)
+                        avio_rb64(in);
+                    if (flags & MOV_TFHD_STSD_ID)
+                        avio_rb32(in);
+                    if (flags & MOV_TFHD_DEFAULT_DURATION)
+                        default_duration = avio_rb32(in);
+                }
+                if (tag == MKBETAG('t', 'r', 'u', 'n')) {
+                    return read_trun_duration(in, default_duration,
+                                              pos + size);
+                }
+                avio_seek(in, pos + size, SEEK_SET);
+            }
+            fprintf(stderr, "Couldn't find trun\n");
+            goto fail;
+        }
+        avio_seek(in, pos + size, SEEK_SET);
+    }
+    fprintf(stderr, "Couldn't find traf\n");
+
+fail:
+    return ret;
 }
 
-static int read_tfra(struct VideoFiles *files, int start_index, AVIOContext *f)
+static int read_tfra(struct Tracks *tracks, int start_index, AVIOContext *f)
 {
     int ret = AVERROR_EOF, track_id;
     int version, fieldlength, i, j;
     int64_t pos   = avio_tell(f);
     uint32_t size = avio_rb32(f);
-    struct VideoFile *vf = NULL;
+    struct Track *track = NULL;
 
     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
         goto fail;
     version = avio_r8(f);
     avio_rb24(f);
     track_id = avio_rb32(f); /* track id */
-    for (i = start_index; i < files->nb_files && !vf; i++)
-        if (files->files[i]->track_id == track_id)
-            vf = files->files[i];
-    if (!vf) {
+    for (i = start_index; i < tracks->nb_tracks && !track; i++)
+        if (tracks->tracks[i]->track_id == track_id)
+            track = tracks->tracks[i];
+    if (!track) {
         /* Ok, continue parsing the next atom */
         ret = 0;
         goto fail;
     }
     fieldlength = avio_rb32(f);
-    vf->chunks  = avio_rb32(f);
-    vf->offsets = av_mallocz(sizeof(*vf->offsets) * vf->chunks);
-    if (!vf->offsets) {
+    track->chunks  = avio_rb32(f);
+    track->offsets = av_mallocz(sizeof(*track->offsets) * track->chunks);
+    if (!track->offsets) {
         ret = AVERROR(ENOMEM);
         goto fail;
     }
-    for (i = 0; i < vf->chunks; i++) {
+    // The duration here is always the difference between consecutive
+    // start times.
+    for (i = 0; i < track->chunks; i++) {
         if (version == 1) {
-            vf->offsets[i].time   = avio_rb64(f);
-            vf->offsets[i].offset = avio_rb64(f);
+            track->offsets[i].time   = avio_rb64(f);
+            track->offsets[i].offset = avio_rb64(f);
         } else {
-            vf->offsets[i].time   = avio_rb32(f);
-            vf->offsets[i].offset = avio_rb32(f);
+            track->offsets[i].time   = avio_rb32(f);
+            track->offsets[i].offset = avio_rb32(f);
         }
         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
             avio_r8(f);
@@ -184,12 +364,44 @@ static int read_tfra(struct VideoFiles *files, int start_index, AVIOContext *f)
         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
             avio_r8(f);
         if (i > 0)
-            vf->offsets[i - 1].duration = vf->offsets[i].time -
-                                          vf->offsets[i - 1].time;
+            track->offsets[i - 1].duration = track->offsets[i].time -
+                                             track->offsets[i - 1].time;
+    }
+    if (track->chunks > 0) {
+        track->offsets[track->chunks - 1].duration = track->offsets[0].time +
+                                                     track->duration -
+                                                     track->offsets[track->chunks - 1].time;
+    }
+    // Now try and read the actual durations from the trun sample data.
+    for (i = 0; i < track->chunks; i++) {
+        int64_t duration = read_moof_duration(f, track->offsets[i].offset);
+        if (duration > 0 && llabs(duration - track->offsets[i].duration) > 3) {
+            // 3 allows for integer duration to drift a few units,
+            // e.g., for 1/3 durations
+            track->offsets[i].duration = duration;
+        }
+    }
+    if (track->chunks > 0) {
+        if (track->offsets[track->chunks - 1].duration <= 0) {
+            fprintf(stderr, "Calculated last chunk duration for track %d "
+                    "was non-positive (%"PRId64"), probably due to missing "
+                    "fragments ", track->track_id,
+                    track->offsets[track->chunks - 1].duration);
+            if (track->chunks > 1) {
+                track->offsets[track->chunks - 1].duration =
+                    track->offsets[track->chunks - 2].duration;
+            } else {
+                track->offsets[track->chunks - 1].duration = 1;
+            }
+            fprintf(stderr, "corrected to %"PRId64"\n",
+                    track->offsets[track->chunks - 1].duration);
+            track->duration = track->offsets[track->chunks - 1].time +
+                              track->offsets[track->chunks - 1].duration -
+                              track->offsets[0].time;
+            fprintf(stderr, "Track duration corrected to %"PRId64"\n",
+                    track->duration);
+        }
     }
-    if (vf->chunks > 0)
-        vf->offsets[vf->chunks - 1].duration = vf->duration -
-                                               vf->offsets[vf->chunks - 1].time;
     ret = 0;
 
 fail:
@@ -197,10 +409,12 @@ fail:
     return ret;
 }
 
-static int read_mfra(struct VideoFiles *files, int start_index,
-                     const char *file, int split)
+static int read_mfra(struct Tracks *tracks, int start_index,
+                     const char *file, int split, int ismf,
+                     const char *basename, const char* output_prefix)
 {
     int err = 0;
+    const char* err_str = "";
     AVIOContext *f = NULL;
     int32_t mfra_size;
 
@@ -209,43 +423,55 @@ static int read_mfra(struct VideoFiles *files, int start_index,
     avio_seek(f, avio_size(f) - 4, SEEK_SET);
     mfra_size = avio_rb32(f);
     avio_seek(f, -mfra_size, SEEK_CUR);
-    if (avio_rb32(f) != mfra_size)
+    if (avio_rb32(f) != mfra_size) {
+        err = AVERROR_INVALIDDATA;
+        err_str = "mfra size mismatch";
         goto fail;
-    if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a'))
+    }
+    if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
+        err = AVERROR_INVALIDDATA;
+        err_str = "mfra tag mismatch";
         goto fail;
-    while (!read_tfra(files, start_index, f)) {
+    }
+    while (!read_tfra(tracks, start_index, f)) {
         /* Empty */
     }
 
-    if (split)
-        write_fragments(files, start_index, f);
+    if (split || ismf)
+        err = write_fragments(tracks, start_index, f, basename, split, ismf,
+                              output_prefix);
+    err_str = "error in write_fragments";
 
 fail:
     if (f)
         avio_close(f);
+    if (err)
+        fprintf(stderr, "Unable to read the MFRA atom in %s (%s)\n", file, err_str);
     return err;
 }
 
-static int get_private_data(struct VideoFile *vf, AVCodecContext *codec)
+static int get_private_data(struct Track *track, AVCodecContext *codec)
 {
-    vf->codec_private_size = codec->extradata_size;
-    vf->codec_private      = av_mallocz(codec->extradata_size);
-    if (!vf->codec_private)
+    track->codec_private_size = codec->extradata_size;
+    track->codec_private      = av_mallocz(codec->extradata_size);
+    if (!track->codec_private)
         return AVERROR(ENOMEM);
-    memcpy(vf->codec_private, codec->extradata, codec->extradata_size);
+    memcpy(track->codec_private, codec->extradata, codec->extradata_size);
     return 0;
 }
 
-static int get_video_private_data(struct VideoFile *vf, AVCodecContext *codec)
+static int get_video_private_data(struct Track *track, AVCodecContext *codec)
 {
     AVIOContext *io = NULL;
     uint16_t sps_size, pps_size;
-    int err = AVERROR(EINVAL);
+    int err;
 
-    if (codec->codec_id == CODEC_ID_VC1)
-        return get_private_data(vf, codec);
+    if (codec->codec_id == AV_CODEC_ID_VC1)
+        return get_private_data(track, codec);
 
-    avio_open_dyn_buf(&io);
+    if ((err = avio_open_dyn_buf(&io)) < 0)
+        goto fail;
+    err = AVERROR(EINVAL);
     if (codec->extradata_size < 11 || codec->extradata[0] != 1)
         goto fail;
     sps_size = AV_RB16(&codec->extradata[6]);
@@ -261,16 +487,18 @@ static int get_video_private_data(struct VideoFile *vf, AVCodecContext *codec)
     err = 0;
 
 fail:
-    vf->codec_private_size = avio_close_dyn_buf(io, &vf->codec_private);
+    track->codec_private_size = avio_close_dyn_buf(io, &track->codec_private);
     return err;
 }
 
-static int handle_file(struct VideoFiles *files, const char *file, int split)
+static int handle_file(struct Tracks *tracks, const char *file, int split,
+                       int ismf, const char *basename,
+                       const char* output_prefix)
 {
     AVFormatContext *ctx = NULL;
-    int err = 0, i, orig_files = files->nb_files;
+    int err = 0, i, orig_tracks = tracks->nb_tracks;
     char errbuf[50], *ptr;
-    struct VideoFile *vf;
+    struct Track *track;
 
     err = avformat_open_input(&ctx, file, NULL, NULL);
     if (err < 0) {
@@ -290,72 +518,92 @@ static int handle_file(struct VideoFiles *files, const char *file, int split)
         fprintf(stderr, "No streams found in %s\n", file);
         goto fail;
     }
-    if (!files->duration)
-        files->duration = ctx->duration;
 
     for (i = 0; i < ctx->nb_streams; i++) {
+        struct Track **temp;
         AVStream *st = ctx->streams[i];
-        vf = av_mallocz(sizeof(*vf));
-        files->files = av_realloc(files->files,
-                                  sizeof(*files->files) * (files->nb_files + 1));
-        files->files[files->nb_files] = vf;
-
-        vf->name = file;
-        if ((ptr = strrchr(file, '/')) != NULL)
-            vf->name = ptr + 1;
-
-        vf->bitrate   = st->codec->bit_rate;
-        vf->track_id  = st->id;
-        vf->timescale = st->time_base.den;
-        vf->duration  = av_rescale_rnd(ctx->duration, vf->timescale,
-                                       AV_TIME_BASE, AV_ROUND_UP);
-        vf->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
-        vf->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
-
-        if (!vf->is_audio && !vf->is_video) {
+
+        if (st->codec->bit_rate == 0) {
+            fprintf(stderr, "Skipping track %d in %s as it has zero bitrate\n",
+                    st->id, file);
+            continue;
+        }
+
+        track = av_mallocz(sizeof(*track));
+        if (!track) {
+            err = AVERROR(ENOMEM);
+            goto fail;
+        }
+        temp = av_realloc(tracks->tracks,
+                          sizeof(*tracks->tracks) * (tracks->nb_tracks + 1));
+        if (!temp) {
+            av_free(track);
+            err = AVERROR(ENOMEM);
+            goto fail;
+        }
+        tracks->tracks = temp;
+        tracks->tracks[tracks->nb_tracks] = track;
+
+        track->name = file;
+        if ((ptr = strrchr(file, '/')))
+            track->name = ptr + 1;
+
+        track->bitrate   = st->codec->bit_rate;
+        track->track_id  = st->id;
+        track->timescale = st->time_base.den;
+        track->duration  = st->duration;
+        track->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
+        track->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
+
+        if (!track->is_audio && !track->is_video) {
             fprintf(stderr,
                     "Track %d in %s is neither video nor audio, skipping\n",
-                    vf->track_id, file);
-            av_freep(&files->files[files->nb_files]);
+                    track->track_id, file);
+            av_freep(&tracks->tracks[tracks->nb_tracks]);
             continue;
         }
 
-        if (vf->is_audio) {
-            if (files->audio_file < 0)
-                files->audio_file = files->nb_files;
-            files->nb_audio_files++;
-            vf->channels    = st->codec->channels;
-            vf->sample_rate = st->codec->sample_rate;
-            if (st->codec->codec_id == CODEC_ID_AAC) {
-                vf->fourcc    = "AACL";
-                vf->tag       = 255;
-                vf->blocksize = 4;
-            } else if (st->codec->codec_id == CODEC_ID_WMAPRO) {
-                vf->fourcc    = "WMAP";
-                vf->tag       = st->codec->codec_tag;
-                vf->blocksize = st->codec->block_align;
+        tracks->duration = FFMAX(tracks->duration,
+                                 av_rescale_rnd(track->duration, AV_TIME_BASE,
+                                                track->timescale, AV_ROUND_UP));
+
+        if (track->is_audio) {
+            if (tracks->audio_track < 0)
+                tracks->audio_track = tracks->nb_tracks;
+            tracks->nb_audio_tracks++;
+            track->channels    = st->codec->channels;
+            track->sample_rate = st->codec->sample_rate;
+            if (st->codec->codec_id == AV_CODEC_ID_AAC) {
+                track->fourcc    = "AACL";
+                track->tag       = 255;
+                track->blocksize = 4;
+            } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
+                track->fourcc    = "WMAP";
+                track->tag       = st->codec->codec_tag;
+                track->blocksize = st->codec->block_align;
             }
-            get_private_data(vf, st->codec);
+            get_private_data(track, st->codec);
         }
-        if (vf->is_video) {
-            if (files->video_file < 0)
-                files->video_file = files->nb_files;
-            files->nb_video_files++;
-            vf->width  = st->codec->width;
-            vf->height = st->codec->height;
-            if (st->codec->codec_id == CODEC_ID_H264)
-                vf->fourcc = "H264";
-            else if (st->codec->codec_id == CODEC_ID_VC1)
-                vf->fourcc = "WVC1";
-            get_video_private_data(vf, st->codec);
+        if (track->is_video) {
+            if (tracks->video_track < 0)
+                tracks->video_track = tracks->nb_tracks;
+            tracks->nb_video_tracks++;
+            track->width  = st->codec->width;
+            track->height = st->codec->height;
+            if (st->codec->codec_id == AV_CODEC_ID_H264)
+                track->fourcc = "H264";
+            else if (st->codec->codec_id == AV_CODEC_ID_VC1)
+                track->fourcc = "WVC1";
+            get_video_private_data(track, st->codec);
         }
 
-        files->nb_files++;
+        tracks->nb_tracks++;
     }
 
     avformat_close_input(&ctx);
 
-    read_mfra(files, orig_files, file, split);
+    err = read_mfra(tracks, orig_tracks, file, split, ismf, basename,
+                    output_prefix);
 
 fail:
     if (ctx)
@@ -363,14 +611,16 @@ fail:
     return err;
 }
 
-static void output_server_manifest(struct VideoFiles *files,
-                                   const char *basename)
+static void output_server_manifest(struct Tracks *tracks, const char *basename,
+                                   const char *output_prefix,
+                                   const char *path_prefix,
+                                   const char *ismc_prefix)
 {
     char filename[1000];
     FILE *out;
     int i;
 
-    snprintf(filename, sizeof(filename), "%s.ism", basename);
+    snprintf(filename, sizeof(filename), "%s%s.ism", output_prefix, basename);
     out = fopen(filename, "w");
     if (!out) {
         perror(filename);
@@ -380,17 +630,17 @@ static void output_server_manifest(struct VideoFiles *files,
     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
     fprintf(out, "\t<head>\n");
     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
-                 "content=\"%s.ismc\" />\n", basename);
+                 "content=\"%s%s.ismc\" />\n", ismc_prefix, basename);
     fprintf(out, "\t</head>\n");
     fprintf(out, "\t<body>\n");
     fprintf(out, "\t\t<switch>\n");
-    for (i = 0; i < files->nb_files; i++) {
-        struct VideoFile *vf = files->files[i];
-        const char *type     = vf->is_video ? "video" : "audio";
-        fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
-                type, vf->name, vf->bitrate);
+    for (i = 0; i < tracks->nb_tracks; i++) {
+        struct Track *track = tracks->tracks[i];
+        const char *type    = track->is_video ? "video" : "audio";
+        fprintf(out, "\t\t\t<%s src=\"%s%s\" systemBitrate=\"%d\">\n",
+                type, path_prefix, track->name, track->bitrate);
         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
-                     "valueType=\"data\" />\n", vf->track_id);
+                     "valueType=\"data\" />\n", track->track_id);
         fprintf(out, "\t\t\t</%s>\n", type);
     }
     fprintf(out, "\t\t</switch>\n");
@@ -399,17 +649,52 @@ static void output_server_manifest(struct VideoFiles *files,
     fclose(out);
 }
 
-static void output_client_manifest(struct VideoFiles *files,
-                                   const char *basename, int split)
+static void print_track_chunks(FILE *out, struct Tracks *tracks, int main,
+                               const char *type)
+{
+    int i, j;
+    int64_t pos = 0;
+    struct Track *track = tracks->tracks[main];
+    int should_print_time_mismatch = 1;
+
+    for (i = 0; i < track->chunks; i++) {
+        for (j = main + 1; j < tracks->nb_tracks; j++) {
+            if (tracks->tracks[j]->is_audio == track->is_audio) {
+                if (track->offsets[i].duration != tracks->tracks[j]->offsets[i].duration) {
+                    fprintf(stderr, "Mismatched duration of %s chunk %d in %s (%d) and %s (%d)\n",
+                            type, i, track->name, main, tracks->tracks[j]->name, j);
+                    should_print_time_mismatch = 1;
+                }
+                if (track->offsets[i].time != tracks->tracks[j]->offsets[i].time) {
+                    if (should_print_time_mismatch)
+                        fprintf(stderr, "Mismatched (start) time of %s chunk %d in %s (%d) and %s (%d)\n",
+                                type, i, track->name, main, tracks->tracks[j]->name, j);
+                    should_print_time_mismatch = 0;
+                }
+            }
+        }
+        fprintf(out, "\t\t<c n=\"%d\" d=\"%"PRId64"\" ",
+                i, track->offsets[i].duration);
+        if (pos != track->offsets[i].time) {
+            fprintf(out, "t=\"%"PRId64"\" ", track->offsets[i].time);
+            pos = track->offsets[i].time;
+        }
+        pos += track->offsets[i].duration;
+        fprintf(out, "/>\n");
+    }
+}
+
+static void output_client_manifest(struct Tracks *tracks, const char *basename,
+                                   const char *output_prefix, int split)
 {
     char filename[1000];
     FILE *out;
     int i, j;
 
     if (split)
-        snprintf(filename, sizeof(filename), "Manifest");
+        snprintf(filename, sizeof(filename), "%sManifest", output_prefix);
     else
-        snprintf(filename, sizeof(filename), "%s.ismc", basename);
+        snprintf(filename, sizeof(filename), "%s%s.ismc", output_prefix, basename);
     out = fopen(filename, "w");
     if (!out) {
         perror(filename);
@@ -417,86 +702,91 @@ static void output_client_manifest(struct VideoFiles *files,
     }
     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
-                 "Duration=\"%"PRId64 "\">\n", files->duration * 10);
-    if (files->video_file >= 0) {
-        struct VideoFile *vf = files->files[files->video_file];
+                 "Duration=\"%"PRId64 "\">\n", tracks->duration * 10);
+    if (tracks->video_track >= 0) {
+        struct Track *track = tracks->tracks[tracks->video_track];
+        struct Track *first_track = track;
         int index = 0;
         fprintf(out,
                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
                 "Chunks=\"%d\" "
                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
-                files->nb_video_files, vf->chunks);
-        for (i = 0; i < files->nb_files; i++) {
-            vf = files->files[i];
-            if (!vf->is_video)
+                tracks->nb_video_tracks, track->chunks);
+        for (i = 0; i < tracks->nb_tracks; i++) {
+            track = tracks->tracks[i];
+            if (!track->is_video)
                 continue;
             fprintf(out,
                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
                     "CodecPrivateData=\"",
-                    index, vf->bitrate, vf->fourcc, vf->width, vf->height);
-            for (j = 0; j < vf->codec_private_size; j++)
-                fprintf(out, "%02X", vf->codec_private[j]);
+                    index, track->bitrate, track->fourcc, track->width, track->height);
+            for (j = 0; j < track->codec_private_size; j++)
+                fprintf(out, "%02X", track->codec_private[j]);
             fprintf(out, "\" />\n");
             index++;
+            if (track->chunks != first_track->chunks)
+                fprintf(stderr, "Mismatched number of video chunks in %s (id: %d, chunks %d) and %s (id: %d, chunks %d)\n",
+                        track->name, track->track_id, track->chunks, first_track->name, first_track->track_id, first_track->chunks);
         }
-        vf = files->files[files->video_file];
-        for (i = 0; i < vf->chunks; i++)
-            fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n", i,
-                    vf->offsets[i].duration);
+        print_track_chunks(out, tracks, tracks->video_track, "video");
         fprintf(out, "\t</StreamIndex>\n");
     }
-    if (files->audio_file >= 0) {
-        struct VideoFile *vf = files->files[files->audio_file];
+    if (tracks->audio_track >= 0) {
+        struct Track *track = tracks->tracks[tracks->audio_track];
+        struct Track *first_track = track;
         int index = 0;
         fprintf(out,
                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
                 "Chunks=\"%d\" "
                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
-                files->nb_audio_files, vf->chunks);
-        for (i = 0; i < files->nb_files; i++) {
-            vf = files->files[i];
-            if (!vf->is_audio)
+                tracks->nb_audio_tracks, track->chunks);
+        for (i = 0; i < tracks->nb_tracks; i++) {
+            track = tracks->tracks[i];
+            if (!track->is_audio)
                 continue;
             fprintf(out,
                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
                     "AudioTag=\"%d\" CodecPrivateData=\"",
-                    index, vf->bitrate, vf->fourcc, vf->sample_rate,
-                    vf->channels, vf->blocksize, vf->tag);
-            for (j = 0; j < vf->codec_private_size; j++)
-                fprintf(out, "%02X", vf->codec_private[j]);
+                    index, track->bitrate, track->fourcc, track->sample_rate,
+                    track->channels, track->blocksize, track->tag);
+            for (j = 0; j < track->codec_private_size; j++)
+                fprintf(out, "%02X", track->codec_private[j]);
             fprintf(out, "\" />\n");
             index++;
+            if (track->chunks != first_track->chunks)
+                fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
+                        track->name, first_track->name);
         }
-        vf = files->files[files->audio_file];
-        for (i = 0; i < vf->chunks; i++)
-            fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
-                    i, vf->offsets[i].duration);
+        print_track_chunks(out, tracks, tracks->audio_track, "audio");
         fprintf(out, "\t</StreamIndex>\n");
     }
     fprintf(out, "</SmoothStreamingMedia>\n");
     fclose(out);
 }
 
-static void clean_files(struct VideoFiles *files)
+static void clean_tracks(struct Tracks *tracks)
 {
     int i;
-    for (i = 0; i < files->nb_files; i++) {
-        av_freep(&files->files[i]->codec_private);
-        av_freep(&files->files[i]->offsets);
-        av_freep(&files->files[i]);
+    for (i = 0; i < tracks->nb_tracks; i++) {
+        av_freep(&tracks->tracks[i]->codec_private);
+        av_freep(&tracks->tracks[i]->offsets);
+        av_freep(&tracks->tracks[i]);
     }
-    av_freep(&files->files);
-    files->nb_files = 0;
+    av_freep(&tracks->tracks);
+    tracks->nb_tracks = 0;
 }
 
 int main(int argc, char **argv)
 {
     const char *basename = NULL;
-    int split = 0, i;
-    struct VideoFiles vf = { 0, .video_file = -1, .audio_file = -1 };
+    const char *path_prefix = "", *ismc_prefix = "";
+    const char *output_prefix = "";
+    char output_prefix_buf[2048];
+    int split = 0, ismf = 0, i;
+    struct Tracks tracks = { 0, .video_track = -1, .audio_track = -1 };
 
     av_register_all();
 
@@ -504,22 +794,43 @@ int main(int argc, char **argv)
         if (!strcmp(argv[i], "-n")) {
             basename = argv[i + 1];
             i++;
+        } else if (!strcmp(argv[i], "-path-prefix")) {
+            path_prefix = argv[i + 1];
+            i++;
+        } else if (!strcmp(argv[i], "-ismc-prefix")) {
+            ismc_prefix = argv[i + 1];
+            i++;
+        } else if (!strcmp(argv[i], "-output")) {
+            output_prefix = argv[i + 1];
+            i++;
+            if (output_prefix[strlen(output_prefix) - 1] != '/') {
+                snprintf(output_prefix_buf, sizeof(output_prefix_buf),
+                         "%s/", output_prefix);
+                output_prefix = output_prefix_buf;
+            }
         } else if (!strcmp(argv[i], "-split")) {
             split = 1;
+        } else if (!strcmp(argv[i], "-ismf")) {
+            ismf = 1;
         } else if (argv[i][0] == '-') {
             return usage(argv[0], 1);
         } else {
-            handle_file(&vf, argv[i], split);
+            if (!basename)
+                ismf = 0;
+            if (handle_file(&tracks, argv[i], split, ismf,
+                            basename, output_prefix))
+                return 1;
         }
     }
-    if (!vf.nb_files || (!basename && !split))
+    if (!tracks.nb_tracks || (!basename && !split))
         return usage(argv[0], 1);
 
     if (!split)
-        output_server_manifest(&vf, basename);
-    output_client_manifest(&vf, basename, split);
+        output_server_manifest(&tracks, basename, output_prefix,
+                               path_prefix, ismc_prefix);
+    output_client_manifest(&tracks, basename, output_prefix, split);
 
-    clean_files(&vf);
+    clean_tracks(&tracks);
 
     return 0;
 }