app.py

import streamlit as st
from transformers import pipeline
import tempfile
import os

# Load Whisper pipeline once
@st.cache_resource
def load_asr_pipeline():
    return pipeline("automatic-speech-recognition", model="openai/whisper-base")

asr = load_asr_pipeline()

st.title("🗣️ WhaleSong: Voice to Text")
st.markdown("Speech coming soon! Upload an audio file. We'll transcribe it using Whisper.")

# Audio input
audio_file = st.file_uploader("🎙️ Upload audio file (wav, mp3, m4a)", type=["wav", "mp3", "m4a"])

if audio_file:
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
        tmp.write(audio_file.read())
        tmp_path = tmp.name

    st.audio(tmp_path)

    with st.spinner("Transcribing..."):
        result = asr(tmp_path)
        st.success("✅ Transcription complete!")
        st.markdown(f"**Transcript:**\n\n{result['text']}")

    os.remove(tmp_path)