Upload 3 files

app.py

@@ -0,0 +1,41 @@
+import gradio as gr
+import joblib
+import torch
+import soundfile as sf
+import numpy as np
+from transformers import pipeline
+
+# 1) Load your trained text classifier
+text_clf = joblib.load("text_pipeline_balanced.joblib")
+
+# 2) Initialize Whisper ASR
+device = 0 if torch.cuda.is_available() else -1
+asr = pipeline(
+    "automatic-speech-recognition",
+    model="openai/whisper-base",
+    chunk_length_s=30,
+    device=device,
+    ignore_warning=True,
+    generate_kwargs={"language":"en","task":"transcribe"}
+)
+
+def classify_audio_file(filepath):
+    audio, sr = sf.read(filepath, dtype="float32")
+    if audio.ndim>1: audio = audio.mean(axis=1)
+    transcript = asr({"array": audio, "sampling_rate": sr})["text"].strip()
+    proba = text_clf.predict_proba([transcript])[0][1]
+    label = "❌ Unsafe" if proba>0.5 else "✅ Safe"
+    return transcript, label, float(proba)
+
+demo = gr.Interface(
+    fn=classify_audio_file,
+    inputs=gr.Audio(source="microphone", type="filepath", label="Record or upload audio"),
+    outputs=[gr.Textbox(label="Transcript"),
+             gr.Label(num_top_classes=2, label="Safety Label"),
+             gr.Number(label="Unsafe Probability")],
+    title="BubbleGuard Audio Safety",
+    description="Transcribe & flag voice-notes as safe/unsafe"
+)
+
+if __name__=="__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,7 @@
+gradio
+transformers
+torch
+soundfile
+joblib
+scikit-learn
+numpy

text_pipeline_balanced.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60bdf7fdefcf512141d90f004907780233c3a110b0909b55e1fc59f6963b85b6
+size 461844