app.py

# pip install ffmpeg ffprobe
import os
import gradio as gr
from transformers import pipeline
from transformers import WhisperProcessor
pipe = pipeline(task="automatic-speech-recognition", model="openai/whisper-tiny")
processor = WhisperProcessor.from_pretrained("openai/whisper-tiny")
pipe.model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="english", task="transcribe")




def recognize(input_audio):
    # print('?????????-------------')
    # print(type(input_audio))
    # print(input_audio)
    output_text = pipe(input_audio)['text']
    return output_text





with gr.Blocks() as demo:
    gr.Markdown("""
                """)
    # 录音功能
    input_audio = gr.Audio(source="microphone", type="filepath", label='Audio')

    audio_button = gr.Button(value="Recognize!")
    text = gr.Textbox(lines=2, label='你要不要看看你说了啥？？？')
    audio_button.click(recognize, inputs=[input_audio], outputs=[text],api_name='predict')


demo.launch()