added the files

app.py

@@ -0,0 +1,76 @@
+import gradio as gr
+from transformers import AutoProcessor, WhisperForConditionalGeneration
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from gtts import gTTS
+import tempfile
+import torch
+
+# 1. Load Whisper STT model (CPU mode)
+processor = AutoProcessor.from_pretrained("openai/whisper-small")
+stt_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-small")
+stt_model.to("cpu")  # Make sure it runs on CPU
+
+# 2. Load TinyLlama (or similar LLM)
+llm_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+tokenizer = AutoTokenizer.from_pretrained(llm_name)
+llm_model = AutoModelForCausalLM.from_pretrained(llm_name)
+llm_model.to("cpu")
+
+# 3. Reference poem for style
+with open("poem_data.txt", "r") as f:
+    reference_poem = f.read().strip()
+
+
+# 4. Transcribe using Whisper
+def transcribe(audio_path):
+    if audio_path is None:
+        return ""
+
+    # Load audio as input features
+    input_features = processor(audio_path, return_tensors="pt", sampling_rate=16000).input_features
+    predicted_ids = stt_model.generate(input_features.to("cpu"))
+    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
+    return transcription.strip()
+
+
+# 5. Generate poem
+def generate_poem(prompt):
+    final_prompt = f"Here is a reference poem:\n{reference_poem}\n\nNow write a new poem about {prompt.strip()} in the same style."
+    inputs = tokenizer.encode(final_prompt, return_tensors="pt", truncation=True)
+    outputs = llm_model.generate(inputs, max_new_tokens=120, temperature=0.7)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+
+# 6. Text-to-speech
+def synthesize(text):
+    tts = gTTS(text)
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as fp:
+        tts.save(fp.name)
+        return fp.name
+
+
+# 7. Gradio pipeline
+def full_pipeline(audio_input, typed_object):
+    obj = typed_object or transcribe(audio_input)
+    poem = generate_poem(obj)
+    audio_poem = synthesize(poem)
+    return poem, audio_poem
+
+
+# 8. Gradio app
+demo = gr.Interface(
+    fn=full_pipeline,
+    inputs=[
+        gr.Audio(source="microphone", type="filepath", label="Speak object"),
+        gr.Textbox(label="Or type object name")
+    ],
+    outputs=[
+        gr.Textbox(label="Generated Poem"),
+        gr.Audio(label="Audio of Poem")
+    ],
+    title="AI Poetry Assistant",
+    description="Speak or type a topic, and the assistant generates a poem in the style of 'A Photograph'."
+)
+
+if __name__ == "__main__":
+    demo.launch()

poem_data.txt

@@ -0,0 +1,19 @@
+The cardboard shows me how it was
+When the two girl cousins went paddling
+Each one holding one of my mother’s hands,
+And she the big girl – some twelve years or so.
+All three stood still to smile through their hair
+At the uncle with the camera, A sweet face
+My mother’s, that was before I was born
+And the sea, which appears to have changed less
+Washed their terribly transient feet.
+Some twenty- thirty- years later
+She’d laugh at the snapshot. “See Betty
+And Dolly,” she’d say, “and look how they
+Dressed us for the beach.” The sea holiday
+was her past, mine is her laughter. Both wry
+With the laboured ease of loss.
+Now she’s has been dead nearly as many years
+As that girl lived. And of this circumstance
+There is nothing to say at all,
+Its silence silences.

requirements.txt

@@ -0,0 +1,7 @@
+transformers
+torch
+gradio
+gtts
+librosa
+ffmpeg-python
+