app.py

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

# Load the pre-trained processor and model
processor = Wav2Vec2Processor.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn")
model = Wav2Vec2ForCTC.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn")

def speech_to_text(audio):
    # Load audio file
    speech, sample_rate = sf.read(audio)

    # Preprocess the audio file
    inputs = processor(speech, sampling_rate=sample_rate, return_tensors="pt", padding=True)

    # Perform inference
    with torch.no_grad():
        logits = model(**inputs).logits

    # Decode the predicted ids to text
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.batch_decode(predicted_ids)

    return transcription[0]

# Create the Gradio interface
iface = gr.Interface(
    fn=speech_to_text,
    inputs=gr.Audio(type="filepath"),
    outputs=gr.Textbox(),
    title="Chinese Speech Recognition",
    description="Upload an audio file and get the transcribed text using the wav2vec2-large-xlsr-53-chinese-zh-cn model."
)

if __name__ == "__main__":
    iface.launch()