app.py

import gradio as gr
import torch
import librosa
from transformers import AutoModelForAudioClassification, Wav2Vec2FeatureExtractor

model_id = "abedir/emotion-detector"

processor = Wav2Vec2FeatureExtractor.from_pretrained(model_id)
model = AutoModelForAudioClassification.from_pretrained(model_id)

label_map = {
    0: "Angry/Fearful",
    1: "Happy/Laugh",
    2: "Neutral/Calm",
    3: "Sad/Cry",
    4: "Surprised/Amazed"
}

def predict(audio):
    audio, sr = librosa.load(audio, sr=16000)
    inputs = processor(audio, sampling_rate=16000, return_tensors="pt")

    with torch.no_grad():
        logits = model(**inputs).logits
        probs = torch.softmax(logits, dim=1)[0]

    pred = torch.argmax(probs).item()
    return label_map[pred], float(probs[pred])

iface = gr.Interface(
    fn=predict,
    inputs=gr.Audio(type="filepath"),
    outputs=["text", "number"],
    title="Emotion Detector 🎤"
)

iface.launch()