Update app.py

app.py

@@ -1,14 +1,15 @@
 from dotenv import find_dotenv, load_dotenv
 from transformers import pipeline
+from transformers import AutoProcessor, AutoModel
+from langchain import PromptTemplate, LLMChain
+from langchain.llms import GooglePalm
+import scipy
 import streamlit as st
-import os
 
-
-# load env variables from .env file
 load_dotenv(find_dotenv())
 
-# img to text
-def img_to_text(url):
+# img2text
+def img_2_text(url):
     image_to_text = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
 
     text = image_to_text(url)[0]["generated_text"]
@@ -16,33 +17,44 @@ def img_to_text(url):
 
 
 # llm
-def generate_story(text):
-    generator = pipeline("text-generation", model="distilgpt2")
+def generate_story(scenario):
+    template = """"
+        You are a story teller;
+        you can generate a creative fun story based on a sample narrative, the story should not be more than 100 words;
+        CONTEXT: {scenario}
+        STORY: 
+        """
+
+    prompt = PromptTemplate(template=template,
+                            input_variables=['scenario']
+                            )
+    llm = GooglePalm(temperature=0.7)
+
+    story_llm = LLMChain(llm=llm, prompt=prompt, verbose=True)
 
-    result = generator(text, max_length=20, num_return_sequences=1)
-    return result[0]['generated_text']
+    story = story_llm.predict(scenario=scenario)
+
+    return story
 
 
 #
 # text-to-speech
 def text_to_speech(text):
-    import requests
-
-    API_URL = "https://api-inference.huggingface.co/models/espnet/kan-bayashi_ljspeech_vits"
-    headers = {"Authorization": f"Bearer {os.environ.get('HUGGINGFACE_API_TOKEN')}"}
-    payload = {
-        "inputs": text
-    }
+    processor = AutoProcessor.from_pretrained("suno/bark-small")
+    model = AutoModel.from_pretrained("suno/bark-small")
 
-    response = requests.post(API_URL, headers=headers, json=payload)
-    response.raise_for_status()
-    with open('audio.flac', 'wb') as file:
-        file.write(response.content)
+    inputs = processor(
+        text=[text],
+        return_tensors="pt",
+    )
 
+    speech_values = model.generate(**inputs, do_sample=True)
+    sampling_rate = model.generation_config.sample_rate
+    scipy.io.wavfile.write("audio.wav", rate=sampling_rate, data=speech_values.cpu().numpy().squeeze())
 
 
 def main():
-    st.set_page_config(page_title="img to audio story")
+    st.set_page_config(page_title="img 2 audio story")
     st.header("turn image to audio story")
     uploaded_file = st.file_uploader("Choose an image ... ", type="jpg")
 
@@ -52,7 +64,7 @@ def main():
         with open(uploaded_file.name, "wb") as file:
             file.write(bytes_data)
         st.image(uploaded_file, caption="Uploaded image", use_column_width=True)
-        text = img_to_text(uploaded_file.name)
+        text = img_2_text(uploaded_file.name)
         story = generate_story(text)
         text_to_speech(story)
 
@@ -60,6 +72,7 @@ def main():
             st.write(text)
         with st.expander("story"):
             st.write(story)
-        st.audio("audio.flac")
+        st.audio("audio.wav")
+
 
 main()