Aesthetics and sampling_rate fix

app.py

@@ -12,6 +12,11 @@ def download_from_youtube(url):
     fpath = streams.first().download()
     return fpath
 
+def get_timestamp(seconds):
+    minutes = int(seconds / 60)
+    seconds = int(seconds % 60)
+    return f"{minutes}:{seconds}"
+
 def divide_into_30s_segments(audio_fpath, seconds_max):
     if not os.path.exists("segmented_audios"):
         os.makedirs("segmented_audios")
@@ -55,12 +60,16 @@ def transcribe(audio, url, seconds_max):
         fpath = download_from_youtube(url)
         segment_paths, segment_start_times = divide_into_30s_segments(fpath, seconds_max)
 
-        audio_dataset = Dataset.from_dict({"audio": segment_paths}).cast_column("audio", Audio())
+        audio_dataset = Dataset.from_dict({"audio": segment_paths}).cast_column("audio", Audio(sampling_rate=16000))
         print(audio_dataset)
         print(audio_dataset[0])
-        text = pipe(audio_dataset["audio"])
+        pred = pipe(audio_dataset["audio"])
         print(type(text))
         print(text)
+        text = ""
+        n_segments = len(segment_start_times)
+        for i, (seconds, output) in enumerate(zip(segment_start_times, pred)):
+            text += f"[Segment {i}/{n_segments}, start time {get_timestamp(seconds)}]\n{output['text']}\n"
         return text
 
     else: