Update app.py

app.py

@@ -1,131 +1,62 @@
-import gradio as gr
 import re
-import requests
-
-# =========================
-# ElevenLabs Configuration
-# =========================
-
-ELEVENLABS_API_KEY = "c92a87a2ebb5f51ee9fe90cc421e836e32780c188f4e0056d77ce69803008ae9"
-
-STT_URL = "https://api.elevenlabs.io/v1/speech-to-text"
-
-
-# =========================
-# Arabic Post Processing
-# =========================
-
-def clean_arabic_text(text):
-    if not text:
-        return ""
-
-    # Remove tashkeel
-    tashkeel_pattern = re.compile(r'[\u0617-\u061A\u064B-\u0652]')
-    text = re.sub(tashkeel_pattern, '', text)
-
-    # Normalize Hamza
-    text = re.sub(r'[أإآ]', 'ا', text)
-
-    # ة → ه
-    text = re.sub(r'ة\b', 'ه', text)
-
-    # ى → ي
-    text = re.sub(r'ى\b', 'ي', text)
-
-    # Remove symbols
-    text = re.sub(r'[^\w\s]', '', text)
-
-    # Remove extra spaces
-    text = " ".join(text.split())
-
-    return text
-
-
-# =========================
-# ElevenLabs Speech To Text
-# =========================
-
-def transcribe_audio(audio_file):
-
-    if audio_file is None:
-        return "No audio uploaded", ""
-
-    headers = {
-        "xi-api-key": ELEVENLABS_API_KEY
-    }
-
-    files = {
-        "file": open(audio_file, "rb")
-    }
-
-    data = {
-        "model_id": "scribe_v2",
-        "enable_logging": "false"
-    }
-
-    response = requests.post(
-        STT_URL,
-        headers=headers,
-        files=files,
-        data=data
-    )
-
-    if response.status_code != 200:
-        return "Error: " + response.text, ""
+import gradio as gr
 
-    result = response.json()
+# regex patterns
+REPEAT_WORD = re.compile(r'\b(\w+)(?:\s+\1\b)+', re.IGNORECASE)
+CHAR_STRETCH = re.compile(r'(.)\1{2,}')
+REPEAT_SYLLABLE = re.compile(r'\b(\w{1,3})(?:\s+\1\b)+', re.IGNORECASE)
 
-    # Extract speaker_0 text
-    text = ""
+def is_filler(word):
+    w = word.lower()
 
-    if "segments" in result:
-        for segment in result["segments"]:
-            if segment.get("speaker") == "speaker_0":
-                text += segment.get("text", "") + " "
+    # حرف واحد مكرر (ممم، ووو، ااا)
+    if len(set(w)) == 1 and len(w) <= 4:
+        return True
 
-    else:
-        text = result.get("text", "")
+    # مقطع قصير جدا
+    if len(w) <= 2:
+        return True
 
-    cleaned = clean_arabic_text(text)
+    return False
 
-    return text, cleaned
 
+def clean_transcript(text):
 
-# =========================
-# Gradio UI
-# =========================
+    # collapse stretched characters
+    text = CHAR_STRETCH.sub(r'\1', text)
 
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    # remove repeated words
+    text = REPEAT_WORD.sub(r'\1', text)
 
-    gr.Markdown("# ElevenLabs Speech To Text + Post Process")
+    # remove repeated short syllables
+    text = REPEAT_SYLLABLE.sub(r'\1', text)
 
-    gr.Markdown(
-        "ارفع ملف صوتي (wav) وسيتم تحويله إلى نص عربي أو إنجليزي مع تنظيف النص."
-    )
+    words = text.split()
 
-    audio_input = gr.Audio(
-        type="filepath",
-        label="Upload audio.wav"
-    )
+    filtered = []
+    for w in words:
+        if not is_filler(w):
+            filtered.append(w)
 
-    raw_text = gr.Textbox(
-        label="Original Text",
-        lines=8
-    )
+    return " ".join(filtered)
 
-    clean_text = gr.Textbox(
-        label="Cleaned Text",
-        lines=8
-    )
 
-    btn = gr.Button("Transcribe")
+def process(text):
+    return clean_transcript(text)
 
-    btn.click(
-        fn=transcribe_audio,
-        inputs=audio_input,
-        outputs=[raw_text, clean_text]
-    )
 
+demo = gr.Interface(
+    fn=process,
+    inputs=gr.Textbox(
+        lines=8,
+        placeholder="Paste transcript here..."
+    ),
+    outputs=gr.Textbox(
+        lines=8,
+        label="Cleaned transcript"
+    ),
+    title="Transcript Filler Cleaner",
+    description="Remove repeated words and speech fillers automatically"
+)
 
-if __name__ == "__main__":
-    demo.launch()
+demo.launch()