fix mp4 stts box bug.

ossrs · Sep 29, 2024 · 58eba6a · 58eba6a
1 parent 40e8ed4
commit 58eba6a
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 17 deletions.
diff --git a/trunk/src/kernel/srs_kernel_mp4.cpp b/trunk/src/kernel/srs_kernel_mp4.cpp
@@ -5062,21 +5062,15 @@ srs_error_t SrsMp4SampleManager::write_track(SrsFrameType track,
         if (sample->frame_type == SrsVideoAvcFrameTypeKeyFrame) {
             stss_entries.push_back(sample->index + 1);
         }
-
-        if (stts) {
-            if (previous) {
-                uint32_t delta = (uint32_t)(sample->dts - previous->dts);
-                if (stts_entry.sample_delta == 0 || stts_entry.sample_delta == delta) {
-                    stts_entry.sample_delta = delta;
-                    stts_entry.sample_count++;
-                } else {
-                    stts_entries.push_back(stts_entry);
-                    stts_entry.sample_count = 1;
-                    stts_entry.sample_delta = delta;
-                }
-            } else {
-                // The first sample always in the STTS table.
-                stts_entry.sample_count++;
+
+        if (stts && previous) {
+            if (sample->dts >= previous->dts && previous->nb_samples > 0) {
+                uint32_t delta = (uint32_t)(sample->dts - previous->dts) / previous->nb_samples;
+                stts_entry.sample_count = previous->nb_samples;
+                // calcaulate delta in the time-scale of the media.
+                // moov->mvhd->timescale which is hardcoded to 1000, sample->tbn also being hardcoded to 1000.
+                stts_entry.sample_delta = delta * previous->tbn / 1000;
+                stts_entries.push_back(stts_entry);
             }
         }
 
@@ -5097,7 +5091,10 @@ srs_error_t SrsMp4SampleManager::write_track(SrsFrameType track,
         previous = sample;
     }
 
-    if (stts && stts_entry.sample_count) {
+    if (stts && previous && previous->nb_samples > 0) {
+        stts_entry.sample_count = previous->nb_samples;
+        // Can't calculate last sample duration, so set sample_delta to 1.
+        stts_entry.sample_delta = 1;
         stts_entries.push_back(stts_entry);
     }
 
@@ -5828,10 +5825,12 @@ srs_error_t SrsMp4Encoder::write_sample(
         ps->type = SrsFrameTypeVideo;
         ps->frame_type = (SrsVideoAvcFrameType)ft;
         ps->index = nb_videos++;
+        ps->nb_samples = format->video->nb_samples;
         vduration = dts;
     } else if (ht == SrsMp4HandlerTypeSOUN) {
         ps->type = SrsFrameTypeAudio;
         ps->index = nb_audios++;
+        ps->nb_samples = format->audio->nb_samples;
         aduration = dts;
     } else {
         srs_freep(ps);

diff --git a/trunk/src/kernel/srs_kernel_mp4.hpp b/trunk/src/kernel/srs_kernel_mp4.hpp
@@ -1554,7 +1554,7 @@ class SrsMp4SampleDescriptionBox : public SrsMp4FullBox
 };
 
 // 8.6.1.2 Decoding Time to Sample Box (stts), for Audio/Video.
-// ISO_IEC_14496-12-base-format-2012.pdf, page 48
+// ISO_IEC_14496-12-base-format-2012.pdf, page 36
 class SrsMp4SttsEntry
 {
 public:
@@ -1893,6 +1893,9 @@ class SrsMp4Sample
     // The sample data.
     uint32_t nb_data;
     uint8_t* data;
+    // number of nalu|audio-frames in this sample.
+    uint32_t nb_samples;
+
 public:
     SrsMp4Sample();
     virtual ~SrsMp4Sample();