app.py

# import whisper
import gradio as gr
import torch

from transformers import pipeline

device = 0 if torch.cuda.is_available() else "cpu"

pipe = pipeline(
    task="automatic-speech-recognition",
    model="openai/whisper-large-v2",
    chunk_length_s=30,
    device=device,
)
pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language="English", task="transcribe")

# model = whisper.load_model("base")

def transcribe(audio):
    
    # audio = whisper.load_audio(audio)
    # audio = whisper.pad_or_trim(audio)

    # result = model.transcribe(audio, verbose = True)

    # mel = whisper.log_mel_spectrogram(audio).to(model.device)

    # options = whisper.DecodingOptions(fp16 = False)
    # result = whisper.decode(model, mel, options)

    # result_text = "Question: " + result.text
    return pipe(audio)["text"]
    
    
 
gr.Interface(
    title = 'Speech to Chat-GPT', 
    fn=transcribe, 
    inputs=[
        gr.inputs.Audio(source="upload", type="filepath")
    ],
    outputs=[
        "textbox"
    ]).launch()