app.py

from pyharp import ModelCard, build_endpoint, save_and_return_filepath
from audiotools import AudioSignal
from audioldm import build_model, text_to_audio
import gradio as gr
import soundfile as sf
from datetime import datetime
import subprocess
import os
import sys

audioldm = build_model(model_name="audioldm-l-full")

def process_fn(input_audio_path, seed, guidance_scale, num_inference_steps, num_candidates, audio_length_in_s):
    video_extensions = (".mp4", ".avi", ".mkv", ".flv", ".mov", ".wmv", ".webm")
    if input_audio_path.lower().endswith(video_extensions):
        input_audio_path = convert_video_to_audio_ffmpeg(input_audio_path)
    
    waveform = text_to_audio(
        audioldm,
        'placeholder', 
        input_audio_path, 
        seed = int(seed),
        duration = audio_length_in_s, 
        guidance_scale = guidance_scale, 
        n_candidate_gen_per_text = int(num_candidates), 
        ddim_steps = int(num_inference_steps)
        )
    
    timestamp = datetime.now().strftime("%Y%m%d%H%M%S")
    filename = f"./ldm_variations_{timestamp}.wav"
    sf.write(filename, waveform[0, 0], samplerate=16000)
    #save_wave(waveform, "./", name="output.wav")
    return filename


def convert_video_to_audio_ffmpeg(video_file, output_ext="wav"):
    """Converts video to audio directly using `ffmpeg` command
    with the help of subprocess module"""
    filename, ext = os.path.splitext(video_file)
    subprocess.call(["ffmpeg", "-y", "-i", video_file, f"{filename}.{output_ext}"], 
                    stdout=subprocess.DEVNULL,
                    stderr=subprocess.STDOUT)
    return f"{filename}.{output_ext}"

webapp = gr.Interface(
    fn = process_fn,
    # Define your Gradio interface
    inputs = [
        gr.Audio(
            label="Audio Input", 
            type="filepath"
        ), 
        gr.Slider(
            label="seed",
            minimum="0",
            maximum="65535",
            value="43534",
            step="1"
        ),
        gr.Slider(
            minimum=0, maximum=10, 
            step=0.1, value=2.5, 
            label="Guidance Scale"
        ),
        gr.Slider(
            minimum=1, maximum=500, 
            step=1, value=200, 
            label="Inference Steps"
        ),
        gr.Slider(
            minimum=1, maximum=10, 
            step=1, value=1, 
            label="Candidates"
        ),
        gr.Slider(
            minimum=2.5, maximum=10.0, 
            step=2.5, value=5, 
            label="Duration"
        )
    ],

    outputs = gr.Audio(label="Audio Output", type="filepath", format="wav", elem_id="audio")
    )

webapp.queue()
webapp.launch()