Create app.py

app.py

@@ -0,0 +1,39 @@
+import gradio as gr
+import torchaudio
+from transformers import AutoModelForSpeechSeq2Seq, PreTrainedTokenizerFast
+
+def transcribe_audio(audio_path):
+    # Load and resample audio
+    audio, sr = torchaudio.load(audio_path)
+    if sr != 16000:
+        audio = torchaudio.functional.resample(audio, sr, 16000)
+    
+    # Get transcription
+    tokens = model(audio)
+    transcription = tokenizer.decode(tokens[0], skip_special_tokens=True)
+    return transcription
+
+# Load model and tokenizer globally
+model = AutoModelForSpeechSeq2Seq.from_pretrained('usefulsensors/moonshine-tiny', trust_remote_code=True)
+tokenizer = PreTrainedTokenizerFast.from_pretrained('usefulsensors/moonshine-tiny')
+
+# Create Gradio interface
+demo = gr.Blocks()
+
+with demo:
+    gr.Markdown("## Audio Transcription App")
+    
+    with gr.Tabs():
+        with gr.TabItem("Upload Audio"):
+            audio_file = gr.Audio(source="upload", type="filepath")
+            output_text1 = gr.Textbox(label="Transcription")
+            upload_button = gr.Button("Transcribe")
+            upload_button.click(fn=transcribe_audio, inputs=audio_file, outputs=output_text1)
+            
+        with gr.TabItem("Record Audio"):
+            audio_mic = gr.Audio(source="microphone", type="filepath")
+            output_text2 = gr.Textbox(label="Transcription")
+            record_button = gr.Button("Transcribe")
+            record_button.click(fn=transcribe_audio, inputs=audio_mic, outputs=output_text2)
+
+demo.launch()