Handle AV_NOPTS_VALUE inputs

ffmpeg/decoder.c

@@ -11,11 +11,43 @@ static int lpms_send_packet(struct input_ctx *ictx, AVCodecContext *dec, AVPacke
     return ret;
 }
 
+static int64_t decoded_video_pts_step(struct input_ctx *ictx, AVFrame *frame)
+{
+    if (frame && frame->duration > 0) return frame->duration;
+    AVStream *vst = (ictx && ictx->ic && ictx->vi >= 0) ? ictx->ic->streams[ictx->vi] : NULL;
+    if (vst && vst->r_frame_rate.num > 0 && vst->r_frame_rate.den > 0) {
+      int64_t step = av_rescale_q(1, av_inv_q(vst->r_frame_rate), vst->time_base);
+      if (step > 0) return step;
+    }
+    return 1;
+}
+
+// Fix malformed decode timestamps (missing/regressive PTS) so downstream stages
+// receive a stable, non-AV_NOPTS_VALUE video timeline.
+static void fix_video_pts(struct input_ctx *ictx, AVFrame *frame)
+{
+    int64_t pts = frame->pts;
+    int synthesized = 0;
+    if (pts == AV_NOPTS_VALUE) pts = frame->best_effort_timestamp;
+    if (pts == AV_NOPTS_VALUE) {
+      pts = decoded_video_pts_step(ictx, frame);
+      if (ictx->last_video_pts != AV_NOPTS_VALUE) pts += ictx->last_video_pts;
+      synthesized = 1;
+    }
+    if (ictx->last_video_pts != AV_NOPTS_VALUE && pts <= ictx->last_video_pts) {
+      int64_t step = synthesized ? decoded_video_pts_step(ictx, frame) : 1;
+      pts = ictx->last_video_pts + step;
+    }
+    frame->pts = pts;
+    ictx->last_video_pts = pts;
+}
+
 static int lpms_receive_frame(struct input_ctx *ictx, AVCodecContext *dec, AVFrame *frame)
 {
     int ret = avcodec_receive_frame(dec, frame);
     if (dec != ictx->vc) return ret;
     if (!ret && frame && !is_flush_frame(frame)) {
+      fix_video_pts(ictx, frame);
       ictx->pkt_diff--; // decrease buffer count for non-sentinel video frames
       if (ictx->flushing) ictx->sentinel_count = 0;
     }
@@ -328,6 +360,7 @@ int open_input(input_params *params, struct input_ctx *ctx)
   if (!ctx->last_frame_v) LPMS_ERR(open_input_err, "Unable to alloc last_frame_v");
   ctx->last_frame_a = av_frame_alloc();
   if (!ctx->last_frame_a) LPMS_ERR(open_input_err, "Unable to alloc last_frame_a");
+  ctx->last_video_pts = AV_NOPTS_VALUE;
 
   return 0;
 

ffmpeg/decoder.h

@@ -36,6 +36,8 @@ struct input_ctx {
 #define SENTINEL_MAX 8
   uint16_t sentinel_count;
 
+  int64_t last_video_pts; // Resets after each segment
+
   // Packet held while decoder is blocked and needs to drain
   AVPacket *blocked_pkt;
 

ffmpeg/ffmpeg_test.go

@@ -2543,8 +2543,8 @@ func TestTranscoder_LargeOutputs(t *testing.T) {
 	close(closeCh)
 	assert.Nil(err)
 	assert.Equal(120, res.Decoded.Frames)
-	assert.Equal(116, res.Encoded[0].Frames) // ffmpeg probably drops missing timestamp frames
+	assert.Equal(120, res.Encoded[0].Frames) // passthrough
-	assert.Equal(56, res.Encoded[1].Frames)
+	assert.Equal(60, res.Encoded[1].Frames)  // 30fps, 2 second input
 	cmd := `
 		# check input properties to ensure they still have the weird timestamps
 		ffprobe -of csv -hide_banner -show_entries frame=pts_time,pkt_dts_time,media_type,pict_type $1/../data/missing-dts.ts 2>&1 | grep video > input.out
@@ -2675,8 +2675,6 @@ func TestTranscoder_LargeOutputs(t *testing.T) {
 
 
 		# check output
-		ls -lha
-		#ffprobe -of csv -hide_banner -show_entries frame=pts_time,pkt_dts_time,media_type,pict_type out-30fps.ts
 		ffprobe -of csv -hide_banner -show_entries frame=pts_time,pkt_dts_time,media_type,pict_type out-30fps.ts 2>&1 | grep video > output.out
 		cat <<- 'EOF2' > expected-output.out
 			frame,video,25994.033333,25994.033333,I,
@@ -2697,9 +2695,10 @@ func TestTranscoder_LargeOutputs(t *testing.T) {
 			frame,video,25994.533333,25994.533333,B
 			frame,video,25994.566667,25994.566667,B
 			frame,video,25994.600000,25994.600000,P,
-			frame,video,25994.666667,25994.666667,P,
+			frame,video,25994.633333,25994.633333,P,
+			frame,video,25994.666667,25994.666667,B,
 			frame,video,25994.700000,25994.700000,B,
-			frame,video,25994.733333,25994.733333,B,
+			frame,video,25994.733333,25994.733333,P,
 			frame,video,25994.766667,25994.766667,B,
 			frame,video,25994.800000,25994.800000,P,
 			frame,video,25994.833333,25994.833333,B,
@@ -2710,30 +2709,33 @@ func TestTranscoder_LargeOutputs(t *testing.T) {
 			frame,video,25995.000000,25995.000000,P,
 			frame,video,25995.033333,25995.033333,B,
 			frame,video,25995.066667,25995.066667,B,
+			frame,video,25995.100000,25995.100000,P,
 			frame,video,25995.133333,25995.133333,B,
 			frame,video,25995.166667,25995.166667,P,
 			frame,video,25995.200000,25995.200000,B,
+			frame,video,25995.233333,25995.233333,B,
 			frame,video,25995.266667,25995.266667,B,
-			frame,video,25995.300000,25995.300000,B,
+			frame,video,25995.300000,25995.300000,P,
-			frame,video,25995.333333,25995.333333,P,
+			frame,video,25995.333333,25995.333333,B,
 			frame,video,25995.366667,25995.366667,B,
 			frame,video,25995.400000,25995.400000,B,
-			frame,video,25995.433333,25995.433333,B,
+			frame,video,25995.433333,25995.433333,P,
-			frame,video,25995.466667,25995.466667,P,
+			frame,video,25995.466667,25995.466667,B,
 			frame,video,25995.500000,25995.500000,B,
 			frame,video,25995.533333,25995.533333,B,
-			frame,video,25995.566667,25995.566667,B,
+			frame,video,25995.566667,25995.566667,P,
-			frame,video,25995.600000,25995.600000,P,
+			frame,video,25995.600000,25995.600000,B,
 			frame,video,25995.633333,25995.633333,B,
 			frame,video,25995.666667,25995.666667,B,
+			frame,video,25995.700000,25995.700000,P,
 			frame,video,25995.733333,25995.733333,B,
-			frame,video,25995.766667,25995.766667,P,
+			frame,video,25995.766667,25995.766667,B,
 			frame,video,25995.800000,25995.800000,B,
-			frame,video,25995.833333,25995.833333,B,
+			frame,video,25995.833333,25995.833333,P,
 			frame,video,25995.866667,25995.866667,B,
-			frame,video,25995.900000,25995.900000,P,
+			frame,video,25995.900000,25995.900000,B,
 			frame,video,25995.933333,25995.933333,B,
-			frame,video,25995.966667,N/A,B,
+			frame,video,25995.966667,N/A,P,
 			frame,video,25996.000000,N/A,P,
 		EOF2
 		diff -u expected-output.out output.out

ffmpeg/transcoder.c

@@ -161,6 +161,7 @@ int transcode_init(struct transcode_thread *h, input_params *inp,
   int nb_outputs = h->nb_outputs;
 
   if (!inp) LPMS_ERR(transcode_cleanup, "Missing input params")
+  ictx->last_video_pts = AV_NOPTS_VALUE;
 
   AVDictionary **demuxer_opts = NULL;
   if (inp->demuxer.opts) demuxer_opts = &inp->demuxer.opts;