Update app.py

app.py

@@ -1,8 +1,7 @@
-import sys, os
 import torch
 from audiocraft.models import MusicGen
 from transformers import GPT2LMHeadModel, GPT2Tokenizer
-from gtts import gTTS
+import pyttsx3
 import gradio as gr
 from tempfile import NamedTemporaryFile
 import numpy as np
@@ -11,150 +10,148 @@ from diffusers import StableDiffusionPipeline
 import matplotlib.pyplot as plt
 import librosa.display
 import librosa
+import soundfile as sf
 from PIL import Image
+import os
 
-# 1) Startup logs
-print("=== STARTUP ===")
-print("Python:", sys.version.replace('\n',' '))
-print("Torch:", torch.__version__)
-print("Device: CPU")
-
-# 2) Force CPU
-device = torch.device("cpu")
-
-# 3) Load MusicGen
-print("Loading MusicGen…")
+# Set device
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# MusicGen
 music_model = MusicGen.get_pretrained("small", device=device)
-print("MusicGen loaded.")
 
-# 4) Load GPT-2
-print("Loading GPT-2…")
+# GPT-2 for conversation
 tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
 gpt2_model = GPT2LMHeadModel.from_pretrained("gpt2").to(device)
-print("GPT-2 loaded.")
 
-# 5) Load Stable Diffusion (CPU-safe)
-print("Loading Stable Diffusion…")
-pipe = StableDiffusionPipeline.from_pretrained(
-    "runwayml/stable-diffusion-v1-5",
-    torch_dtype=torch.float32
-).to(device)
-print("Stable Diffusion loaded.")
+# Stable Diffusion for image generation
+pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", torch_dtype=torch.float16)
+pipe = pipe.to(device)
 
-# Emotion helper
+# Emotion detection for Text-to-Audio
 def get_emotion_tone(text):
-    txt = text.lower()
-    if any(w in txt for w in ["happy","joy","excited"]): return "happy"
-    if any(w in txt for w in ["sad","down","melancholy"]): return "sad"
-    if any(w in txt for w in ["angry","frustrated"]): return "angry"
-    return "neutral"
+    if any(word in text.lower() for word in ["happy", "joy", "excited"]):
+        return "happy"
+    elif any(word in text.lower() for word in ["sad", "down", "melancholy"]):
+        return "sad"
+    elif any(word in text.lower() for word in ["angry", "frustrated"]):
+        return "angry"
+    else:
+        return "neutral"
 
-# Image generation
+# Image generation using Stable Diffusion
 def generate_image(prompt, style="realistic"):
-    styled = f"{style} style {prompt}"
+    styled_prompt = f"{style} style {prompt}"
     try:
-        img = pipe(styled).images[0]
-        tmp = NamedTemporaryFile(delete=False, suffix=".png")
-        img.save(tmp.name)
-        return tmp.name
+        image = pipe(styled_prompt).images[0]
+        temp_image = NamedTemporaryFile(delete=False, suffix=".png")
+        image.save(temp_image.name)
+        return temp_image.name
     except Exception as e:
-        print("Image error:", e)
-        return None
+        return f"Error generating image: {e}"
 
-# Text-to-audio via gTTS
+# Convert Text to Audio with Emotion
 def text_to_audio(text):
-    try:
-        tts = gTTS(text=text, lang="en")
-        tmp = NamedTemporaryFile(delete=False, suffix=".mp3")
-        tts.save(tmp.name)
-        return tmp.name
-    except Exception as e:
-        print("TTS error:", e)
-        return None
+    emotion = get_emotion_tone(text)
+    engine = pyttsx3.init()
+    engine.setProperty('rate', 150 if emotion == "neutral" else 180 if emotion == "happy" else 100 if emotion == "sad" else 200)
+    engine.setProperty('volume', 0.8 if emotion == "neutral" else 1.0 if emotion == "happy" or emotion == "angry" else 0.5)
 
-# Music generation
+    temp_file = NamedTemporaryFile(delete=False, suffix=".mp3")
+    engine.save_to_file(text, temp_file.name)
+    engine.runAndWait()
+    return temp_file.name
+
+# Music generation using MusicGen
 def generate_music(prompt):
     try:
-        wav = music_model.generate([prompt])  # [1,1,T]
-        data = wav.cpu().numpy()[0,0]
-        tmp = NamedTemporaryFile(delete=False, suffix=".wav")
-        wavfile.write(tmp.name, music_model.sample_rate, data)
-        return tmp.name
+        descriptions = [prompt]
+        wav = music_model.generate(descriptions)
+        temp_file = NamedTemporaryFile(delete=False, suffix=".wav")
+        audio_data = wav.cpu().numpy()
+        wavfile.write(temp_file.name, music_model.sample_rate, audio_data[0, 0])
+        return temp_file.name
     except Exception as e:
-        print("Music error:", e)
-        return None
+        return f"Error generating music: {e}"
 
-# Spectrogram
+# Spectrogram generation from audio
 def generate_spectrogram(audio_path):
     try:
-        y,sr = librosa.load(audio_path, sr=None)
+        y, sr = librosa.load(audio_path, sr=None)
         S = librosa.feature.melspectrogram(y, sr=sr)
-        S_db = librosa.power_to_db(S, ref=np.max)
-        plt.figure(figsize=(6,3))
-        librosa.display.specshow(S_db, sr=sr, x_axis='time', y_axis='mel')
-        tmp = NamedTemporaryFile(delete=False, suffix=".png")
-        plt.savefig(tmp.name, bbox_inches="tight")
+        S_dB = librosa.power_to_db(S, ref=np.max)
+
+        plt.figure(figsize=(10, 4))
+        librosa.display.specshow(S_dB, sr=sr, x_axis='time', y_axis='mel', cmap='coolwarm')
+        plt.colorbar(format='%+2.0f dB')
+        plt.title('Mel-frequency spectrogram')
+        temp_image = NamedTemporaryFile(delete=False, suffix=".png")
+        plt.savefig(temp_image.name)
         plt.close()
-        return tmp.name
+        return temp_image.name
     except Exception as e:
-        print("Spectrogram error:", e)
-        return None
+        return f"Error generating spectrogram: {e}"
 
-# GPT-2 chat
-def chat_with_ai(text):
+# Chat with AI (GPT-2)
+def chat_with_ai(user_input):
     try:
-        tok = tokenizer.encode(text, return_tensors="pt").to(device)
-        out = gpt2_model.generate(tok, max_length=50)
-        return tokenizer.decode(out[0], skip_special_tokens=True)
+        inputs = tokenizer.encode(user_input, return_tensors="pt").to(device)
+        outputs = gpt2_model.generate(inputs, max_length=50, num_return_sequences=1)
+        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        return response
     except Exception as e:
-        print("Chat error:", e)
-        return "Error generating response."
+        return f"Error in chat generation: {e}"
 
-# GIF video
+# Simulate Video Generation using a Sequence of Images
 def generate_video(prompt):
-    frames=[]
-    for i in range(5):
-        p = generate_image(f"{prompt} frame {i+1}")
-        if p: frames.append(Image.open(p))
-    if not frames: return None
-    tmp = NamedTemporaryFile(delete=False, suffix=".gif")
-    frames[0].save(tmp.name, save_all=True, append_images=frames[1:], duration=400, loop=0)
-    return tmp.name
-
-# Main interface
-def main(input_text, task, style):
-    if task=="Conversation":
-        resp = chat_with_ai(input_text)
-        img  = generate_image(f"conversation about {input_text}", style)
-        return resp, None, img
-    if task=="Music":
-        mus  = generate_music(input_text)
-        spec = generate_spectrogram(mus) if mus else None
-        return "Music ready", mus, spec
-    if task=="Text to Audio":
-        aud = text_to_audio(input_text)
-        img = generate_image(f"audio for {input_text}", style)
-        return "Audio ready", aud, img
-    if task=="Video Generation":
-        vid = generate_video(input_text)
-        aud = generate_music(input_text)
-        return "Video ready", aud, vid
-
-# Launch with debug logging
-iface = gr.Interface(
-    fn=main,
+    frames = []
+    for i in range(5):  # Generate 5 frames as a sequence
+        frame_prompt = f"{prompt} frame {i+1}"
+        frame_path = generate_image(frame_prompt)
+        frames.append(Image.open(frame_path))
+
+    temp_video = NamedTemporaryFile(delete=False, suffix=".gif")
+    frames[0].save(temp_video.name, save_all=True, append_images=frames[1:], duration=500, loop=0)
+    return temp_video.name
+
+# Main interface logic
+def main_interface(input_text, task_type, style):
+    try:
+        if task_type == "Conversation":
+            response = chat_with_ai(input_text)
+            image_path = generate_image(f"conversation about {input_text}", style)
+            return response, None, image_path
+
+        elif task_type == "Music":
+            audio_path = generate_music(input_text)
+            spectrogram_path = generate_spectrogram(audio_path)
+            return "Music Generated", audio_path, spectrogram_path
+
+        elif task_type == "Text to Audio":
+            audio_path = text_to_audio(input_text)
+            image_path = generate_image(f"text-to-audio conversion for {input_text}", style)
+            return "Audio Generated", audio_path, image_path
+
+        elif task_type == "Video Generation":
+            video_path = generate_video(input_text)
+            audio_path = generate_music(input_text)
+            return "Video Generated", audio_path, video_path
+    except Exception as e:
+        return f"Error: {e}", None, None
+
+# Gradio interface setup
+interface = gr.Interface(
+    fn=main_interface,
     inputs=[
-        gr.Textbox(label="Enter Prompt"),
-        gr.Radio(["Conversation","Music","Text to Audio","Video Generation"], label="Task"),
-        gr.Dropdown(["realistic","abstract","comic"], label="Style"),
+        gr.Textbox(label="Enter Text or Prompt"),
+        gr.Radio(["Conversation", "Music", "Text to Audio", "Video Generation"], label="Select Task"),
+        gr.Dropdown(["realistic", "abstract", "comic"], label="Select Style"),
     ],
     outputs=[
-        gr.Textbox(label="Output Text"),
-        gr.Audio(label="Audio File", type="filepath"),
-        gr.Image(label="Image/GIF", type="filepath"),
+        gr.Textbox(label="Generated Output"),
+        gr.Audio(label="Generated Audio", type="filepath"),
+        gr.Image(label="Generated Image", type="filepath"),
     ],
+    live=False,
 )
 
-if __name__=="__main__":
-    print("Launching Gradio…")
-    iface.launch(server_name="0.0.0.0", server_port=7860, debug=True)
+interface.launch()