Update app.py

app.py

@@ -4,9 +4,9 @@ import uuid
 import re
 from pydub import AudioSegment, silence
 
-# -----------------------------
-# Filename cleaner
-# -----------------------------
+# -------------------------------------------------
+# Clean filename
+# -------------------------------------------------
 def clean_file_name(file_path):
     base = os.path.basename(file_path)
     name, ext = os.path.splitext(base)
@@ -17,18 +17,21 @@ def clean_file_name(file_path):
     uid = uuid.uuid4().hex[:6]
     return os.path.join(os.path.dirname(file_path), f"{name}_{uid}{ext}")
 
-# -----------------------------
-# Fast silence removal
-# -----------------------------
-def remove_silence_fast(file_path, keep_silence_ms=50):
+# -------------------------------------------------
+# SAFE Silence Removal (NO WORD LOSS)
+# -------------------------------------------------
+def remove_silence_safe(file_path, keep_silence_ms=250):
     audio = AudioSegment.from_file(file_path)
+
+    # mono for speed, safe for voice
     audio = audio.set_channels(1)
 
-    silence_thresh = audio.dBFS - 16
+    # 🔐 conservative threshold (speech friendly)
+    silence_thresh = audio.dBFS - 22
 
     chunks = silence.split_on_silence(
         audio,
-        min_silence_len=120,
+        min_silence_len=350,       # ignore natural speech pauses
         silence_thresh=silence_thresh,
         keep_silence=keep_silence_ms
     )
@@ -36,60 +39,81 @@ def remove_silence_fast(file_path, keep_silence_ms=50):
     if not chunks:
         return file_path
 
-    out = chunks[0]
-    for c in chunks[1:]:
-        out += c
+    output = AudioSegment.empty()
+
+    for chunk in chunks:
+        # soften edges so words never cut
+        chunk = chunk.fade_in(20).fade_out(20)
+        output += chunk
 
-    out_path = clean_file_name(file_path)
-    out.export(out_path)
-    return out_path
+    output_path = clean_file_name(file_path)
+    output.export(output_path)
+    return output_path
 
-# -----------------------------
-# Duration
-# -----------------------------
+# -------------------------------------------------
+# Duration helper
+# -------------------------------------------------
 def duration(path):
-    return len(AudioSegment.from_file(path)) / 1000
+    audio = AudioSegment.from_file(path)
+    return len(audio) / 1000
 
-# -----------------------------
-# Main function
-# -----------------------------
-def process_audio(audio_file, seconds):
-    keep_ms = int(seconds * 1000)
+# -------------------------------------------------
+# Main processing
+# -------------------------------------------------
+def process_audio(audio_file, keep_seconds):
+    keep_ms = int(max(keep_seconds, 0.25) * 1000)
 
     before = duration(audio_file)
-    output = remove_silence_fast(audio_file, keep_ms)
+    output = remove_silence_safe(audio_file, keep_ms)
     after = duration(output)
 
-    text = (
-        f"Old Duration : {before:.2f}s\n"
-        f"New Duration : {after:.2f}s\n"
-        f"Saved        : {before - after:.2f}s"
+    info = (
+        f"Old Duration : {before:.2f} sec\n"
+        f"New Duration : {after:.2f} sec\n"
+        f"Silence Removed : {before - after:.2f} sec"
     )
 
-    return output, output, text
-
-# -----------------------------
-# UI
-# -----------------------------
-with gr.Blocks(title="Remove Silence From Audio") as demo:
-    gr.Markdown("## 🔇 Fast Audio Silence Remover")
+    return output, output, info
+
+# -------------------------------------------------
+# UI (Hugging Face SAFE)
+# -------------------------------------------------
+with gr.Blocks(title="Safe Silence Remover") as demo:
+    gr.Markdown(
+        """
+        ## 🔇 Safe Silence Remover  
+        ✔ Removes **only silence**  
+        ✔ Words & voice **100% preserved**  
+        """
+    )
 
     with gr.Row():
         with gr.Column():
-            audio = gr.Audio(type="filepath", sources=["upload", "microphone"])
-            silence_keep = gr.Number(value=0.05, label="Keep Silence (seconds)")
+            audio_input = gr.Audio(
+                label="Upload Audio",
+                type="filepath",
+                sources=["upload", "microphone"]
+            )
+
+            keep_silence = gr.Number(
+                label="Keep Silence (seconds)",
+                value=0.25
+            )
+
             btn = gr.Button("Remove Silence")
 
         with gr.Column():
-            out_audio = gr.Audio()
-            out_file = gr.File()
-            info = gr.Textbox()
+            audio_out = gr.Audio(label="Processed Audio")
+            file_out = gr.File(label="Download")
+            info_out = gr.Textbox(label="Info")
 
     btn.click(
         process_audio,
-        [audio, silence_keep],
-        [out_audio, out_file, info]
+        [audio_input, keep_silence],
+        [audio_out, file_out, info_out]
     )
 
-# ✅ HF EXPECTS THIS ONLY
+# -------------------------------------------------
+# Hugging Face Entry Point
+# -------------------------------------------------
 demo.queue().launch()