diff --git a/dali/operators/reader/video_reader_decoder_op_test.cc b/dali/operators/reader/video_reader_decoder_op_test.cc
index 6112fc1aafe..89d412963a0 100644
--- a/dali/operators/reader/video_reader_decoder_op_test.cc
+++ b/dali/operators/reader/video_reader_decoder_op_test.cc
@@ -124,18 +124,23 @@ class VideoReaderDecoderBaseTest : public VideoTestBase {
     auto &ground_truth_video = cfr_videos_[0];
 
     Pipeline pipe(batch_size, 1, device_id, seed);
+    std::vector<std::int> labels(cfr_videos_[0].NumFrames());
+    std::iota(a.begin(), a.end(), 1);
 
     pipe.AddOperator(OpSpec("experimental__readers__Video")
       .AddArg("device", backend)
       .AddArg("sequence_length", sequence_length)
       .AddArg("random_shuffle", true)
       .AddArg("initial_fill", cfr_videos_[0].NumFrames())
+      .AddArg(
+        "labels", labels)
       .AddArg(
         "filenames",
         std::vector<std::string>{cfr_videos_paths_[0]})
-      .AddOutput("frames", backend));
+      .AddOutput("frames", backend)
+      .AddOutput("frame_labels", backend));
 
-    pipe.Build({{"frames", backend}});
+    pipe.Build({{"frames", backend}, {"frame_labels", backend}});
 
     // ToFix
     std::vector<int> expected_order = {29, 46, 33, 6, 37};
@@ -148,7 +153,12 @@ class VideoReaderDecoderBaseTest : public VideoTestBase {
       pipe.Outputs(&ws);
 
       auto &frame_video_output = ws.Output<Backend>(0);
+      auto &cpu_label_output = ws.Output<Backend(1);
       const auto sample = frame_video_output.template tensor<uint8_t>(0);
+      const auto cpu_label = cpu_label_output.tensor<int>(sample_id);
+      int label_buffer = -1;
+      MemCopy(&label_buffer, cpu_label, sizeof(int));
+      std::cout << " labels " << label_buffer << " vs " << expected_order[sequence_id],
 
       // We want to access correct order, so we compare only the first frame of the sequence
       AssertFrame(expected_order[sequence_id], sample, ground_truth_video);