diff --git a/jasper/data/asr_recycler.py b/jasper/data/asr_recycler.py
index cc38f8b..57e0bc6 100644
--- a/jasper/data/asr_recycler.py
+++ b/jasper/data/asr_recycler.py
@@ -59,10 +59,13 @@ def extract_data(
             code_wav = code_fb.getvalue()
             # only starting 1 min audio has reliable alignment ignore rest
             if start_time > 60:
+                print(f'start time over 60 seconds of audio skipping.')
                 break
             # only if some reasonable audio data is present yield it
-            if code_seg.duration_seconds >= 0.5:
-                yield transcript, code_seg.duration_seconds, code_wav
+            if code_seg.duration_seconds < 0.5:
+                print(f'transcript chunk "{transcript}" contains no audio skipping.')
+                continue
+            yield transcript, code_seg.duration_seconds, code_wav
 
     def asr_data_generator(call_wav, call_wav_fname, events):
         call_wav_0, call_wav_1 = call_wav.split_to_mono()
diff --git a/jasper/data/rev_recycler.py b/jasper/data/rev_recycler.py
index 9e45bbd..c83204a 100644
--- a/jasper/data/rev_recycler.py
+++ b/jasper/data/rev_recycler.py
@@ -91,8 +91,10 @@ def extract_data(
             text = "".join(lens["elements"].Each()["value"].collect()(monologue))
             text_clean = re.sub(r"\[.*\]", "", text)
             # only if some reasonable audio data is present yield it
-            if tscript_wav_seg.duration_seconds >= 0.5:
-                yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav
+            if tscript_wav_seg.duration_seconds < 0.5:
+                print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.')
+                continue
+            yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav
 
     def mono_asr_data_generator(wav_seg, wav_path, meta):
         monologues = lens["monologues"].Each().collect()(meta)
@@ -122,6 +124,9 @@ def extract_data(
             tscript_wav = tscript_wav_fb.getvalue()
             text = "".join(lens["elements"].Each()["value"].collect()(monologue))
             text_clean = re.sub(r"\[.*\]", "", text)
+            if tscript_wav_seg.duration_seconds < 0.5:
+                print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.')
+                continue
             yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav
 
     def generate_rev_asr_data():