Update app.py

app.py

@@ -3,6 +3,8 @@ import torch
 import streamlit as st
 from groq import Groq
 from diffusers import AutoPipelineForText2Image
+import tempfile
+import soundfile as sf
 
 # Load API keys
 GROQ_API_KEY = os.getenv("GROQ_API_KEY")
@@ -11,21 +13,17 @@ HF_API_KEY = os.getenv("HF_API_KEY")
 # Initialize Groq client with API key
 client = Groq(api_key=GROQ_API_KEY)
 
-# Select device (GPU if available, else CPU)
-device = "cuda" if torch.cuda.is_available() else "cpu"
-st.write(f"Using device: {device}")  # Display device info
-
 # Load lightweight Hugging Face image generation model
 image_gen = AutoPipelineForText2Image.from_pretrained(
     "stabilityai/sdxl-turbo", use_auth_token=HF_API_KEY
 )
-image_gen.to(device)
+image_gen.to("cuda" if torch.cuda.is_available() else "cpu")
 
 # Function to transcribe Tamil audio using Groq's Whisper
-def transcribe(audio_file):
-    with open(audio_file, "rb") as file:
+def transcribe(audio_path):
+    with open(audio_path, "rb") as file:
         transcription = client.audio.transcriptions.create(
-            file=(audio_file, file.read()),
+            file=(audio_path, file.read()),
             model="whisper-large-v3",
             language="ta",  # Tamil
             response_format="verbose_json"
@@ -38,7 +36,7 @@ def translate_text(tamil_text):
         model="gemma-7b-it",
         messages=[{"role": "user", "content": f"Translate this Tamil text to English: {tamil_text}"}]
     )
-    return response.choices[0].message.content
+    return response.choices[0].delta.content
 
 # Function to generate text using Groq's DeepSeek R1
 def generate_text(prompt):
@@ -46,7 +44,7 @@ def generate_text(prompt):
         model="deepseek-coder-r1-7b",
         messages=[{"role": "user", "content": f"Write a short story about: {prompt}"}]
     )
-    return response.choices[0].message.content
+    return response.choices[0].delta.content
 
 # Function to generate an image
 def generate_image(prompt):
@@ -56,29 +54,41 @@ def generate_image(prompt):
 # Streamlit UI
 st.title("Tamil Speech to Image & Story Generator")
 
-# File uploader for audio
-uploaded_audio = st.file_uploader("Upload your Tamil speech", type=["wav", "mp3", "m4a"])
-
-if uploaded_audio is not None:
-    st.audio(uploaded_audio, format="audio/wav")
-    
-    if st.button("Generate"):
-        with st.spinner("Transcribing..."):
-            tamil_text = transcribe(uploaded_audio)
-            st.success("Transcription complete!")
-            st.text_area("Tamil Text Output", tamil_text)
-        
-        with st.spinner("Translating to English..."):
-            english_text = translate_text(tamil_text)
-            st.success("Translation complete!")
-            st.text_area("Translated English Text", english_text)
-        
-        with st.spinner("Generating story..."):
-            story = generate_text(english_text)
-            st.success("Story generation complete!")
-            st.text_area("Generated Story", story)
-        
-        with st.spinner("Generating image..."):
-            image = generate_image(english_text)
-            st.success("Image generation complete!")
-            st.image(image, caption="Generated Image")
+# Audio input - Recording or Uploading
+st.subheader("Upload or Record Audio")
+recorded_audio = st.audio("", format='audio/wav', start_time=0)
+uploaded_file = st.file_uploader("Upload an audio file", type=["wav", "mp3", "m4a"]) 
+
+audio_path = None
+
+if uploaded_file is not None:
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
+        temp_audio.write(uploaded_file.read())
+        audio_path = temp_audio.name
+elif recorded_audio:
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
+        audio_data, samplerate = sf.read(recorded_audio)
+        sf.write(temp_audio.name, audio_data, samplerate)
+        audio_path = temp_audio.name
+
+if st.button("Generate") and audio_path:
+    with st.spinner("Transcribing Tamil speech..."):
+        tamil_text = transcribe(audio_path)
+    with st.spinner("Translating to English..."):
+        english_text = translate_text(tamil_text)
+    with st.spinner("Generating story..."):
+        story = generate_text(english_text)
+    with st.spinner("Generating image..."):
+        image = generate_image(english_text)
+
+    st.subheader("Tamil Transcription")
+    st.write(tamil_text)
+
+    st.subheader("English Translation")
+    st.write(english_text)
+
+    st.subheader("Generated Story")
+    st.write(story)
+
+    st.subheader("Generated Image")
+    st.image(image)