app.py

import gradio as gr
import torch
import soundfile as sf
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

# Load model
processor = Wav2Vec2Processor.from_pretrained("nguyenvulebinh/wav2vec2-base-vietnamese-250h")
model = Wav2Vec2ForCTC.from_pretrained("nguyenvulebinh/wav2vec2-base-vietnamese-250h")

def transcribe(audio):
    # Load audio
    speech, rate = sf.read(audio)
    if rate != 16000:
        return "Vui lòng cung cấp file audio 16kHz."

    # Preprocess and predict
    inputs = processor(speech, sampling_rate=16000, return_tensors="pt", padding=True)
    with torch.no_grad():
        logits = model(**inputs).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.decode(predicted_ids[0])
    return transcription

# Gradio UI
gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(type="filepath", label="Upload audio (16kHz, mono)"),
    outputs="text",
    title="Wav2Vec2 Vietnamese STT",
    description="Nhận dạng giọng nói tiếng Việt bằng mô hình wav2vec2-base-vietnamese-250h từ VLSP."
).launch()