app.py

import torch
import gradio as gr
from transformers import pipeline

MODEL_NAME = "openai/whisper-small"  # this always needs to stay in line 8 :D sorry for the hackiness
lang = "en"

device = 0 if torch.cuda.is_available() else "cpu"
pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    chunk_length_s=30,
    device=device,
)

pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language=lang, task="transcribe")

def transcribe(microphone, file_upload):
    warn_output = ""
    if microphone and file_upload:
        warn_output = (
            "WARNING: You've uploaded an audio file and used the microphone. "
            "The recorded file from the microphone will be used, and the uploaded audio will be discarded.\n"
        )

    elif not (microphone or file_upload):
        return "ERROR: You have to either use the microphone or upload an audio file."

    file = microphone if microphone else file_upload
    text = pipe(file)["text"]
    return warn_output + text

examples = [
    ['Martin Luther king - FREE AT LAST.mp3'], 
    ['Winston Churchul - ARCH OF VICTOR.mp3'], 
    ['Voice of Neil Armstrong.mp3'], 
    ['Speeh by George Washington.mp3'], 
    ['Speech by John Kennedy.mp3'], 
    ['Al Gore on Inventing the Internet.mp3'], 
    ['Alan Greenspan.mp3'], 
    ['Neil Armstrong - ONE SMALL STEP.mp3'], 
    ['General Eisenhower announcing D-Day landing.mp3'], 
    ['Hey Siri.wav']
]

css = """
footer {display:none !important}
.output-markdown{display:none !important}
button.primary {
    z-index: 14;
    left: 0px;
    top: 0px;
    cursor: pointer !important; 
    background: none rgb(17, 20, 45) !important;
    border: none !important;
    color: rgb(255, 255, 255) !important;
    line-height: 1 !important;
    border-radius: 6px !important;
    transition: box-shadow 200ms ease 0s, background 200ms ease 0s !important;
    box-shadow: none !important;
}
button.primary:hover{
    z-index: 14;
    left: 0px;
    top: 0px;
    cursor: pointer !important;
    background: none rgb(66, 133, 244) !important;
    border: none !important;
    color: rgb(255, 255, 255) !important;
    line-height: 1 !important;
    border-radius: 6px !important;
    transition: box-shadow 200ms ease 0s, background 200ms ease 0s !important;
    box-shadow: rgb(0 0 0 / 23%) 0px 1px 7px 0px !important;
}
button.gallery-item:hover {
    border-color: rgb(37 56 133) !important;
    background-color: rgb(229,225,255) !important;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Row():
        gr.Markdown("## Speech Recognition Demo")
    with gr.Row():
        mic_input = gr.Audio(label="Microphone Input", interactive=True, type="filepath")
        file_upload = gr.Audio(label="File Upload", interactive=True, type="filepath")
    with gr.Row():
        output = gr.Textbox(label="Transcription Output")
    with gr.Row():
        gr.Examples(examples=examples, inputs=[file_upload], label="Examples")

    transcribe_button = gr.Button("Transcribe")
    transcribe_button.click(transcribe, inputs=[mic_input, file_upload], outputs=[output])

demo.launch()