Update app.py

app.py

@@ -11,11 +11,12 @@ MUSICA_FONDO = "musica.mp3"
 TAG1 = "tag.mp3"
 TAG2 = "tag2.mp3"
 
-def mezclar_musica_y_tags(audio_path: str) -> str:
+def mezclar_musica_y_tags(audio_path: str, custom_music_path: str = None) -> str:
     podcast_audio = AudioSegment.from_file(audio_path)
-    musica_fondo = AudioSegment.from_file(MUSICA_FONDO).apply_gain(-15)
-    tag1_audio = AudioSegment.from_file(TAG1).apply_gain(-5)
-    tag2_audio = AudioSegment.from_file(TAG2).apply_gain(-5)
+    music_file = custom_music_path if custom_music_path and os.path.exists(custom_music_path) else MUSICA_FONDO
+    musica_fondo = AudioSegment.from_file(music_file).apply_gain(-15)
+    tag_outro = AudioSegment.from_file(TAG1).apply_gain(-5)
+    tag_trans = AudioSegment.from_file(TAG2).apply_gain(-5)
 
     duracion_podcast = len(podcast_audio)
     repeticiones = (duracion_podcast // len(musica_fondo)) + 1
@@ -23,21 +24,30 @@ def mezclar_musica_y_tags(audio_path: str) -> str:
     musica_fondo_loop = musica_fondo_loop[:duracion_podcast]
 
     mezcla = musica_fondo_loop.overlay(podcast_audio)
-    mezcla = tag1_audio + mezcla + tag2_audio
+    mezcla = mezcla + tag_outro  # tag.mp3 como outro
+
+    silent_ranges = []
+    for i in range(0, len(podcast_audio) - 500, 100):
+        chunk = podcast_audio[i:i+500]
+        if chunk.dBFS < -40:
+            silent_ranges.append((i, i + 500))
+    for start, end in reversed(silent_ranges):
+        if (end - start) >= len(tag_trans):
+            mezcla = mezcla.overlay(tag_trans, position=start + 50)
 
     output_path = audio_path.replace(".mp3", "_con_musica.mp3")
     mezcla.export(output_path, format="mp3")
     return output_path
 
-def synthesize_sync(article_url, text_input, language, skip_llm, agregar_musica):
-    return asyncio.run(synthesize(article_url, text_input, language, skip_llm, agregar_musica))
+def synthesize_sync(article_url, text_input, language, skip_llm, agregar_musica, custom_music, custom_prompt):
+    return asyncio.run(synthesize(article_url, text_input, language, skip_llm, agregar_musica, custom_music, custom_prompt))
 
-async def synthesize(article_url, text_input, language="en", skip_llm=False, agregar_musica=False):
+async def synthesize(article_url, text_input, language="en", skip_llm=False, agregar_musica=False, custom_music=None, custom_prompt=None):
     if not article_url and not text_input:
         return "Error: Ingresa una URL o texto", None
 
     try:
-        config = ConversationConfig()
+        config = ConversationConfig(custom_prompt_template=custom_prompt)
         converter = URLToAudioConverter(config, llm_api_key=os.environ.get("TOGETHER_API_KEY"))
         
         voices = {
@@ -47,14 +57,14 @@ async def synthesize(article_url, text_input, language="en", skip_llm=False, agr
         voice1, voice2 = voices.get(language, voices["en"])
 
         if skip_llm and text_input:
-            output_file, conversation = await converter.raw_text_to_audio(text_input, voice1, voice2)
+            output_file, conversation = await converter.raw_text_to_audio(text_input, voice1, voice2, custom_music)
         elif text_input:
-            output_file, conversation = await converter.text_to_audio(text_input, voice1, voice2)
+            output_file, conversation = await converter.text_to_audio(text_input, voice1, voice2, custom_music)
         else:
-            output_file, conversation = await converter.url_to_audio(article_url, voice1, voice2)
+            output_file, conversation = await converter.url_to_audio(article_url, voice1, voice2, custom_music)
 
         if agregar_musica:
-            output_file = mezclar_musica_y_tags(output_file)
+            output_file = mezclar_musica_y_tags(output_file, custom_music)
 
         return conversation, output_file
     except Exception as e:
@@ -68,6 +78,11 @@ with gr.Blocks(theme='gstaff/sketch') as demo:
         language = gr.Dropdown(["en", "es"], label="Idioma", value="en")
         skip_llm = gr.Checkbox(label="🔴 Modo libre (sin filtros LLM)", value=False)
         agregar_musica = gr.Checkbox(label="🎵 Agregar música de fondo y cortinillas", value=False)
+        custom_music = gr.File(label="Subir música de fondo (opcional)", file_types=[".mp3"])
+        custom_prompt = gr.Textbox(
+            label="Prompt personalizado (opcional)",
+            placeholder="{text}\nCrea un diálogo de podcast en español entre Anfitrión1 y Anfitrión2. Usa un tono informal y genera al menos 6 intercambios por hablante. Devuelve SOLO un objeto JSON: {\"conversation\": [{\"speaker\": \"Anfitrión1\", \"text\": \"...\"}, {\"speaker\": \"Anfitrión2\", \"text\": \"...\"}]}"
+        )
         btn = gr.Button("Generar Podcast", variant="primary")
     
     with gr.Row():
@@ -76,8 +91,8 @@ with gr.Blocks(theme='gstaff/sketch') as demo:
     
     btn.click(
         synthesize_sync,
-        inputs=[text_url, text_input, language, skip_llm, agregar_musica],
+        inputs=[text_url, text_input, language, skip_llm, agregar_musica, custom_music, custom_prompt],
         outputs=[conv_display, aud]
     )
 
-demo.launch()
+demo.launch()