app.py

import torch
import time

import gradio as gr
import spaces
from transformers import pipeline
from transformers.pipelines.audio_utils import ffmpeg_read

DEFAULT_MODEL_NAME = "openai/whisper-tiny"

BATCH_SIZE = 8

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

def load_pipeline(model_name):
    return pipeline(
        task="automatic-speech-recognition",
        model=model_name,
        chunk_length_s=30,
        device=device,
    )

pipe = load_pipeline(DEFAULT_MODEL_NAME)

@spaces.GPU
def transcribe(inputs, task, model_name):
    if inputs is None:
        raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")

    global pipe
    if model_name != pipe.model.name_or_path:
        pipe = load_pipeline(model_name)

    start_time = time.time()  # Record the start time
    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
    end_time = time.time()  # Record the end time

    transcription_time = end_time - start_time  # Calculate the transcription time

    # Create the transcription time output with additional information
    transcription_time_output = (
        f"Transcription Time: {transcription_time:.2f} seconds\n"
        f"Model Used: {model_name}\n"
        f"Device Used: {'GPU' if torch.cuda.is_available() else 'CPU'}"
    )

    return text, transcription_time_output

demo = gr.Blocks()

mf_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(type="filepath"),
        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
        gr.Textbox(
            label="Model Name",
            value=DEFAULT_MODEL_NAME,
            placeholder="Enter the model name",
            info="Some available models: distil-whisper/distil-large-v3 distil-whisper/distil-medium.en   Systran/faster-distil-whisper-large-v3 Systran/faster-whisper-large-v3  Systran/faster-whisper-medium    openai/whisper-tiny , openai/whisper-base, openai/whisper-medium, openai/whisper-large-v3"
        ),
    ],
    outputs=[gr.TextArea(label="Transcription"), gr.TextArea(label="Transcription Info")],
    theme="huggingface",
    title="Whisper Transcription",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the specified OpenAI Whisper"
        " checkpoint and 🤗 Transformers to transcribe audio files of arbitrary length."
    ),
    allow_flagging="never",
)

file_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(type="filepath", label="Audio file"),
        gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
        gr.Textbox(
            label="Model Name",
            value=DEFAULT_MODEL_NAME,
            placeholder="Enter the model name",
            info="Some available models: openai/whisper-tiny, openai/whisper-base, openai/whisper-medium, openai/whisper-large-v2"
        ),
    ],
    outputs=[gr.TextArea(label="Transcription"), gr.TextArea(label="Transcription Info")],
    theme="huggingface",
    title="Whisper Transcription",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the specified OpenAI Whisper"
        " checkpoint and 🤗 Transformers to transcribe audio files of arbitrary length."
    ),
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface([mf_transcribe, file_transcribe], ["Microphone", "Audio file"])

demo.launch(share=True)