app.py

import gradio as gr
import os
from moviepy.editor import AudioFileClip, ImageClip
from pathlib import Path
import subprocess
import uuid
import shutil


output_dir = Path("temp/")
output_dir.mkdir(exist_ok=True, parents=True)

os.chdir(output_dir) # change working directory to output_dir because the hf spaces model has no option to specify output directory ¯\_(ツ)_/¯


class SpotifyApi:
    spotify_directory = Path(output_dir / "spotify")
    final_directory = output_dir

    def __init__(self):
        self.setup_spotipy

    def setup_spotipy(self) -> None:
        # Check if the credentials file exists
        if not os.path.exists("spotif.rc"):
            with open("spotify.rc", "w") as f:
                f.write(
                    f"{os.environ['SPOTIFY_USERNAME']} {os.environ['SPOTIFY_PASSWORD']}"
                )
            subprocess.call(["spodcast", "-l", "spotify.rc"])
        else:
            pass

    def download_episode(self, episode_url) -> str:
        # Generate a 8 character random string
        foldername = str(uuid.uuid4())[:8]
        out_path = (self.spotify_directory / foldername).resolve()
        subprocess.call(["spodcast", "--root-path", out_path, episode_url])
        self.foldername = foldername
        mp3_path = self.get_final_mp3()
        assert mp3_path is not None
        return mp3_path

    def get_final_mp3(self) -> str | None:
        # Look in all the subdirectories of spotify for the mp3 file
        for root, dirs, files in os.walk(
            Path(self.spotify_directory / self.foldername)
        ):
            for file in files:
                if file.endswith(".mp3"):
                    final_mp3 = Path(
                        self.final_directory / self.foldername
                    ).with_suffix(".mp3")
                    shutil.rmtree(Path(self.spotify_directory / self.foldername))
                    shutil.copy(os.path.join(root, file), final_mp3)
                    return final_mp3.as_posix()


def process_inputs(prompt, audio, spotify_url):
    image = get_stable_diffusion_image(prompt)
    if spotify_url:
        spotify = SpotifyApi()
        audio = spotify.download_episode(spotify_url)
    video = add_static_image_to_audio(image, audio)
    return video


def add_static_image_to_audio(image, audio) -> str:
    """Create and save a video file to `output_path` after
    combining a static image that is located in `image_path`
    with an audio file in `audio_path`"""
    audio_clip = AudioFileClip(audio)
    image_clip = ImageClip(image)
    video_clip = image_clip.set_audio(audio_clip)
    video_clip.duration = audio_clip.duration
    video_clip.fps = 1
    path = Path(output_dir / "output.mp4")
    video_clip.write_videofile(path)
    return path.as_posix()


def get_stable_diffusion_image(prompt) -> str:
    stable_diffusion = gr.Blocks.load(name="spaces/stabilityai/stable-diffusion")
    gallery_dir = stable_diffusion(prompt, fn_index=2)
    return [os.path.join(gallery_dir, img) for img in os.listdir(gallery_dir)][0]


iface = gr.Interface(
    fn=process_inputs,
    inputs=[
        gr.Textbox(label="Describe your podcast clip"),
        gr.Audio(type="filepath", label="Upload an mp3"),
        gr.Textbox(label="Or Paste a spotify episode link"),
    ],
    outputs="video",
)


iface.launch()