app.py

from transformers import pipeline, Wav2Vec2ForCTC, Wav2Vec2Processor
import gradio as gr
import torch
import numpy as np

# Load models
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h").to("cpu")
classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

labels = [
    "secret agent", "lecturer", "motivational speaker", "comedian",
    "philosopher", "calm leader", "emotional artist", "enthusiastic teacher",
    "strict officer", "mysterious thinker"
]

# Transcription + personality
def transcribe_and_analyze(audio):
    if audio is None:
        return "No audio provided", "No analysis"

    # Mono audio
    if len(audio.shape) > 1:
        audio = np.mean(audio, axis=1)
    
    inputs = processor(audio, sampling_rate=16000, return_tensors="pt")
    input_values = inputs.input_values
    with torch.no_grad():
        logits = model(input_values).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.decode(predicted_ids[0]).lower()

    result = classifier(transcription, candidate_labels=labels)
    top_label = result["labels"][0]
    score = result["scores"][0]

    return transcription, f"You sound like a **{top_label}** (confidence: {score:.2f})"

# Gradio UI
with gr.Blocks(theme=gr.themes.Soft()) as app:
    gr.Markdown("# Voice2PersonaAI (Free AI-Based Personality Engine)")

    with gr.Row():
        audio_input = gr.Audio(sources=["microphone", "upload"], type="numpy", label="🎤 Your Voice")

    transcribe_button = gr.Button("Transcribe & Analyze")

    with gr.Row():
        transcribed_text = gr.Textbox(label="Transcription")
        persona_text = gr.Textbox(label="AI-Predicted Persona")

    transcribe_button.click(transcribe_and_analyze, inputs=audio_input, outputs=[transcribed_text, persona_text])

app.launch()