Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 from transformers import pipeline
-
+import librosa
 # ----------------
 # TEXT MODELS
 # ----------------
@@ -42,9 +42,11 @@ segmentation_pipeline = pipeline(
 
 def segment_image(image):
     results = segmentation_pipeline(image)
-    # Gradio AnnotatedImage expects (image, annotations)
-    ann = [(image, r["mask"]) for r in results]
-    return (image, ann)
+    # Combine masks into a single image with labels
+    annotated = {}
+    for r in results:
+        annotated[r["label"]] = r["mask"]  # label → mask
+    return (image, annotated)
 
 
 # ----------------
@@ -52,10 +54,13 @@ def segment_image(image):
 # ----------------
 asr_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
 
-def transcribe(audio):
-    return asr_pipeline(audio)["text"]
 
 
+def transcribe(audio):
+    # Load with max 30s duration
+    speech, sr = librosa.load(audio, sr=16000, duration=30)
+    return asr_pipeline({"array": speech, "sampling_rate": sr}, return_timestamps=True)["text"]
+
 # ----------------
 # GRADIO APP
 # ----------------