Update spaces.py

spaces.py

@@ -31,11 +31,8 @@ def get_voices():
     response = requests.get(url, headers=headers)
     if response.status_code == 200:
         voices = response.json()["voices"]
-        voice_options = {voice["name"]: voice["voice_id"] for voice in voices}
-        return voice_options
-    else:
-        print(f"Error getting voices: {response.status_code}")
-        return {}
+        return [(voice["name"], voice["voice_id"]) for voice in voices]
+    return []
 
 def get_convo_list(description):
     prompt =f"Your task is to return a JSON object representing a complete conversation containing a key 'turns' with a value which is just a list of objects containing 'turn_number', an integer, and 'message', the message for that turn. Ensure you return as many turns as the user specifies, if they specify. Remember, each turn is a turn in a conversation between a phone agent (female) and a human (male). The phone agent should speak first. The conversation is described as:\n{description}.\nCritically, ensure that the human turns employ filler words (uh, uhhhh, ummmm, yeahhh, hm, hmm, etc with repeated letters to denote thinking...) and realistic language without using *sounds effects*. I repeat, do NOT use *sound effects*. Additionally, do not over-use filler words or start every human response with them. The goal is to sound realistic, not exagerrated. The AI should be conversational, employing transition phrases. The AI should always end their response with a question except when saying goodbye. Additionally, digits spaced out. For instance, the human might say: 'My phone number is 8 3 1... 5 4 8... 9 2 2 3...' instead of writing it out. They might also say 'My email is steve at gmail dot com.' where it is written out. Now provide the JSON."
@@ -173,10 +170,10 @@ def create_video_clip(image, duration, target_resolution=(1920, 1080)):
     return clip.set_duration(duration)
 
 def process_message(args):
-    i, message, logo_image, female_voice_id, male_voice_id, male_stability, male_style = args
-    voice_id = female_voice_id if i % 2 == 0 else male_voice_id
+    i, message, logo_image, voice_ids, male_stability, male_style = args
+    voice_id = voice_ids[i % len(voice_ids)]
 
-    if i % 2 == 0:
+    if i % len(voice_ids) == 0:
         text_color = "#cdfa8a"
         stability = 0.8
         style = 0
@@ -211,7 +208,7 @@ def process_message(args):
         print(f"Error processing message {i+1}: {e}")
         return (None, None, None)
 
-def generate_conversation_video(messages, female_voice_id, male_voice_id, logo_url, male_stability, male_style):
+def generate_conversation_video(messages, voice_ids, logo_url, male_stability, male_style):
     logo_image = download_and_convert_svg_to_png(logo_url)
     if logo_image is None:
         return None
@@ -220,7 +217,7 @@ def generate_conversation_video(messages, female_voice_id, male_voice_id, logo_u
     audio_clips = []
     temp_audio_paths = []
 
-    args = [(i, message, logo_image, female_voice_id, male_voice_id, male_stability, male_style) for i, message in enumerate(messages)]
+    args = [(i, message, logo_image, voice_ids, male_stability, male_style) for i, message in enumerate(messages)]
     max_workers = 5
 
     with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
@@ -274,20 +271,31 @@ def generate_conversation_video(messages, female_voice_id, male_voice_id, logo_u
     return temp_video_path
 
 def generate_video(description, female_voice, male_voice, male_stability=0.65, male_style=0.35):
-    voice_options = get_voices()
-    female_voice_id = voice_options[female_voice]
-    male_voice_id = voice_options[male_voice]
-    
+    voice_ids = [
+        female_voice,  # First speaker (female)
+        male_voice     # Second speaker (male)
+    ]
     logo_url = "https://opencall.ai/images/logo-symbol.svg"
     
     messages = get_convo_list(description)
-    video_path = generate_conversation_video(messages, female_voice_id, male_voice_id, logo_url, male_stability, male_style)
+    video_path = generate_conversation_video(messages, voice_ids, logo_url, male_stability, male_style)
     
     return video_path
 
-# Create Gradio interface
-voice_options = get_voices()
+# Get available voices
+voices = get_voices()
+default_female_id = "cgSgspJ2msm6clMCkdW9"  # Default female voice ID
+default_male_id = "3Niy6MUaDzcs7Liw7dFs"    # Default male voice ID
 
+# Create voice selection dropdowns
+female_voice_names = [(voice[0], voice[1]) for voice in voices]
+male_voice_names = [(voice[0], voice[1]) for voice in voices]
+
+# Set default selections
+default_female_idx = next((i for i, v in enumerate(female_voice_names) if v[1] == default_female_id), 0)
+default_male_idx = next((i for i, v in enumerate(male_voice_names) if v[1] == default_male_id), 0)
+
+# Create Gradio interface
 iface = gr.Interface(
     fn=generate_video,
     inputs=[
@@ -298,13 +306,17 @@ iface = gr.Interface(
             info="You can be specific about the number of turns, tone, and content of the conversation"
         ),
         gr.Dropdown(
-            choices=list(voice_options.keys()),
+            choices=female_voice_names,
+            value=female_voice_names[default_female_idx][1],
             label="Female Voice",
+            type="value",
             info="Select the voice for the phone agent"
         ),
         gr.Dropdown(
-            choices=list(voice_options.keys()),
-            label="Male Voice", 
+            choices=male_voice_names,
+            value=male_voice_names[default_male_idx][1],
+            label="Male Voice",
+            type="value",
             info="Select the voice for the customer"
         ),
         gr.Slider(