Update app.py

app.py

@@ -135,6 +135,30 @@ def load_chatterbox_model():
     
     return False
 
+def download_audio_from_url(url):
+    """Download audio from URL and save to temporary file"""
+    try:
+        logger.info(f"📥 Downloading reference audio from: {url}")
+        response = requests.get(url, timeout=30, headers={
+            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'
+        })
+        
+        if response.status_code == 200:
+            # Create temporary file
+            temp_file = tempfile.NamedTemporaryFile(suffix=".flac", delete=False)
+            temp_file.write(response.content)
+            temp_file.close()
+            
+            logger.info(f"✅ Audio downloaded to: {temp_file.name}")
+            return temp_file.name
+        else:
+            logger.error(f"❌ HTTP {response.status_code} when downloading audio")
+            return None
+            
+    except Exception as e:
+        logger.error(f"❌ Error downloading audio from URL: {e}")
+        return None
+
 def get_or_load_model():
     """Load ChatterboxTTS model if not already loaded"""
     global MODEL
@@ -259,10 +283,29 @@ def generate_tts_audio(
         logger.warning("🚨 USING FALLBACK - Real ChatterboxTTS not found!")
         logger.warning("📋 To fix: Upload your ChatterboxTTS package to this Space")
     
+    # Handle audio prompt - download if it's a URL
+    audio_prompt_path = audio_prompt_path_input
+    temp_audio_file = None
+    
     try:
+        # Check if it's a URL
+        if audio_prompt_path_input and (audio_prompt_path_input.startswith('http://') or audio_prompt_path_input.startswith('https://')):
+            logger.info(f"🌐 Detected URL, downloading audio: {audio_prompt_path_input}")
+            temp_audio_file = download_audio_from_url(audio_prompt_path_input)
+            if temp_audio_file:
+                audio_prompt_path = temp_audio_file
+                logger.info(f"✅ Using downloaded audio: {audio_prompt_path}")
+            else:
+                logger.warning("⚠️ Failed to download audio, proceeding without reference")
+                audio_prompt_path = None
+        elif audio_prompt_path_input and not os.path.exists(audio_prompt_path_input):
+            logger.warning(f"⚠️ Audio file not found: {audio_prompt_path_input}, proceeding without reference")
+            audio_prompt_path = None
+        
+        # Generate audio
         wav = current_model.generate(
             text_input[:300],  # Limit text length
-            audio_prompt_path=audio_prompt_path_input,
+            audio_prompt_path=audio_prompt_path,
             exaggeration=exaggeration_input,
             temperature=temperature_input,
             cfg_weight=cfgw_input,
@@ -278,6 +321,14 @@ def generate_tts_audio(
     except Exception as e:
         logger.error(f"❌ Audio generation failed: {e}")
         raise
+    finally:
+        # Clean up temporary file
+        if temp_audio_file and os.path.exists(temp_audio_file):
+            try:
+                os.unlink(temp_audio_file)
+                logger.info(f"🗑️ Cleaned up temporary file: {temp_audio_file}")
+            except:
+                pass
 
 # FastAPI app for API endpoints
 app = FastAPI(
@@ -341,47 +392,68 @@ async def synthesize_speech(request: TTSRequest):
         
         start_time = time.time()
         
-        # Generate audio
-        sample_rate, audio_data = generate_tts_audio(
-            request.text,
-            request.audio_prompt_url,
-            request.exaggeration,
-            request.temperature,
-            request.seed,
-            request.cfg_weight
-        )
+        # Handle audio prompt URL
+        audio_prompt_path = request.audio_prompt_url
+        temp_audio_file = None
         
-        generation_time = time.time() - start_time
-        
-        # Save audio file
-        audio_id = generate_id()
-        audio_path = os.path.join(AUDIO_DIR, f"{audio_id}.wav")
-        sf.write(audio_path, audio_data, sample_rate)
-        
-        # Cache audio info
-        audio_cache[audio_id] = {
-            "path": audio_path,
-            "text": request.text,
-            "sample_rate": sample_rate,
-            "duration": len(audio_data) / sample_rate,
-            "generated_at": time.time(),
-            "generation_time": generation_time,
-            "real_chatterbox": CHATTERBOX_AVAILABLE
-        }
+        if request.audio_prompt_url and (request.audio_prompt_url.startswith('http://') or request.audio_prompt_url.startswith('https://')):
+            temp_audio_file = download_audio_from_url(request.audio_prompt_url)
+            if temp_audio_file:
+                audio_prompt_path = temp_audio_file
+            else:
+                logger.warning("Failed to download reference audio, proceeding without")
+                audio_prompt_path = None
         
-        message = "Speech synthesized successfully"
-        if not CHATTERBOX_AVAILABLE:
-            message += " (using fallback - upload ChatterboxTTS for real synthesis)"
-        
-        logger.info(f"✅ Audio saved: {audio_id} ({generation_time:.2f}s)")
-        
-        return TTSResponse(
-            success=True,
-            audio_id=audio_id,
-            message=message,
-            sample_rate=sample_rate,
-            duration=len(audio_data) / sample_rate
-        )
+        try:
+            # Generate audio
+            sample_rate, audio_data = generate_tts_audio(
+                request.text,
+                audio_prompt_path,
+                request.exaggeration,
+                request.temperature,
+                request.seed,
+                request.cfg_weight
+            )
+            
+            generation_time = time.time() - start_time
+            
+            # Save audio file
+            audio_id = generate_id()
+            audio_path = os.path.join(AUDIO_DIR, f"{audio_id}.wav")
+            sf.write(audio_path, audio_data, sample_rate)
+            
+            # Cache audio info
+            audio_cache[audio_id] = {
+                "path": audio_path,
+                "text": request.text,
+                "sample_rate": sample_rate,
+                "duration": len(audio_data) / sample_rate,
+                "generated_at": time.time(),
+                "generation_time": generation_time,
+                "real_chatterbox": CHATTERBOX_AVAILABLE
+            }
+            
+            message = "Speech synthesized successfully"
+            if not CHATTERBOX_AVAILABLE:
+                message += " (using fallback - upload ChatterboxTTS for real synthesis)"
+            
+            logger.info(f"✅ Audio saved: {audio_id} ({generation_time:.2f}s)")
+            
+            return TTSResponse(
+                success=True,
+                audio_id=audio_id,
+                message=message,
+                sample_rate=sample_rate,
+                duration=len(audio_data) / sample_rate
+            )
+        
+        finally:
+            # Clean up temporary audio file
+            if temp_audio_file and os.path.exists(temp_audio_file):
+                try:
+                    os.unlink(temp_audio_file)
+                except:
+                    pass
         
     except HTTPException:
         raise
@@ -501,8 +573,9 @@ def create_gradio_interface():
                 
                 audio_prompt = gr.Textbox(
                     value="https://storage.googleapis.com/chatterbox-demo-samples/prompts/female_shadowheart4.flac",
-                    label="Reference Audio URL",
-                    placeholder="URL to reference audio file"
+                    label="Reference Audio URL or File Path",
+                    placeholder="https://example.com/audio.wav or /path/to/local/file.wav",
+                    info="URL will be downloaded automatically, or use local file path"
                 )
                 
                 with gr.Row():