app.py

import gradio as gr
import requests
import random
import os
import zipfile # built in module for unzipping files (thank god)
import librosa
import time
from infer_rvc_python import BaseLoader
from pydub import AudioSegment

# fucking dogshit toggle
try:
    import spaces
    spaces_status = True
except ImportError:
    spaces_status = False

converter = BaseLoader(only_cpu=False, hubert_path=None, rmvpe_path=None) # <- yeah so like this handles rvc

global pth_file
global index_file

pth_file = "model.pth"
index_file = "model.index"

#CONFIGS
TEMP_DIR = "temp"
MODEL_PREFIX = "model"
PITCH_ALGO_OPT = [
    "pm",
    "harvest",
    "crepe",
    "rmvpe",
    "rmvpe+",
]


os.makedirs(TEMP_DIR, exist_ok=True)

def unzip_file(file):
    filename = os.path.basename(file).split(".")[0] # converts "model.zip" to "model" so we can do things
    with zipfile.ZipFile(file, 'r') as zip_ref:
        zip_ref.extractall(os.path.join(TEMP_DIR, filename)) # might not be very ram efficient...
    return True
    

def progress_bar(total, current): # best progress bar ever trust me sunglasses emoji 😎 
    return "[" + "=" * int(current / total * 20) + ">" + " " * (20 - int(current / total * 20)) + "] " + str(int(current / total * 100)) + "%"

def download_from_url(url, filename=None):
    if "huggingface" not in url:
        return ["The URL must be from huggingface", "Failed", "Failed"]
    if filename is None:
        filename = os.path.join(TEMP_DIR, MODEL_PREFIX + str(random.randint(1, 1000)) + ".zip")
    response = requests.get(url)
    total = int(response.headers.get('content-length', 0)) # bytes to download (length of the file)
    if total > 500000000:

        return ["The file is too large. You can only download files up to 500 MB in size.", "Failed", "Failed"]
    current = 0
    with open(filename, "wb") as f:
        for data in response.iter_content(chunk_size=4096): # download in chunks of 4096 bytes (4kb - helps with memory usage and speed)
            f.write(data)
            current += len(data)
            print(progress_bar(total, current), end="\r") # \r is a carriage return, it moves the cursor to the start of the line so its like tqdm sunglasses emoji 😎
    
    # unzip because the model is in a zip file lel

    try:
        unzip_file(filename)
    except Exception as e:
        return ["Failed to unzip the file", "Failed", "Failed"] # return early if it fails and like tell the user but its dogshit hahahahahahaha 😎 According to all known laws aviation, there is no way a bee should be able to fly.
    unzipped_dir = os.path.join(TEMP_DIR, os.path.basename(filename).split(".")[0]) # just do what we did in unzip_file because we need the directory
    pth_files = []
    index_files = []
    for root, dirs, files in os.walk(unzipped_dir): # could be done more efficiently because nobody stores models in subdirectories but like who cares (it's a futureproofing thing lel)
        for file in files:
            if file.endswith(".pth"):
                pth_files.append(os.path.join(root, file))
            elif file.endswith(".index"):
                index_files.append(os.path.join(root, file))
    
    print(pth_files, index_files) # debug print because im fucking stupid and i need to see what is going on
    global pth_file
    global index_file
    pth_file = pth_files[0]
    index_file = index_files[0]
    return ["Downloaded as " + filename, pth_files[0], index_files[0]]

if spaces_status:
    @spaces.GPU()
    def convert_now(audio_files, random_tag, converter):
        return converter(
            audio_files,
            random_tag,
            overwrite=False,
            parallel_workers=8
        )
else:
    def convert_now(audio_files, random_tag, converter):
        return converter(
            audio_files,
            random_tag,
            overwrite=False,
            parallel_workers=8
        )
    
def run(
    audio_files,
    file_m,
    pitch_alg,
    pitch_lvl,
    file_index,
    index_inf,
    r_m_f,
    e_r,
    c_b_p,
):
    if not audio_files:
        raise ValueError("The audio pls")

    if isinstance(audio_files, str):
        audio_files = [audio_files]

    try:
        duration_base = librosa.get_duration(filename=audio_files[0])
        print("Duration:", duration_base)
    except Exception as e:
        print(e)

    random_tag = "USER_"+str(random.randint(10000000, 99999999))

    converter.apply_conf(
        tag=random_tag,
        file_model=file_m,
        pitch_algo=pitch_alg,
        pitch_lvl=pitch_lvl,
        file_index=file_index,
        index_influence=index_inf,
        respiration_median_filtering=r_m_f,
        envelope_ratio=e_r,
        consonant_breath_protection=c_b_p,
        resample_sr=44100 if audio_files[0].endswith('.mp3') else 0, 
    )
    time.sleep(0.1)

    result = convert_now(audio_files, random_tag, converter)

    return result[0]

with gr.Blocks() as demo:
    gr.Markdown("## Ilaria RVC 💖")
    with gr.Tab("Inference"):
        sound_gui = gr.Audio(value=None,type="filepath",autoplay=False,visible=True,)
        pth_file_ui = gr.Textbox(label="Model pth file",value=pth_file,visible=False,interactive=False,) # gradio is fucking weird (im with stupid v)
        index_file_ui = gr.Textbox(label="Index pth file",value=index_file,visible=False,interactive=False,) # gradio is fucking weird (im with stupid ^)
        pitch_algo_conf = gr.Dropdown(PITCH_ALGO_OPT,value=PITCH_ALGO_OPT[4],label="Pitch algorithm",visible=True,interactive=True,)
        pitch_lvl_conf = gr.Slider(label="Pitch level",minimum=-24,maximum=24,step=1,value=0,visible=True,interactive=True,)
        index_inf_conf =  gr.Slider(minimum=0,maximum=1,label="Index influence",value=0.75,)
        respiration_filter_conf = gr.Slider(minimum=0,maximum=7,label="Respiration median filtering",value=3,step=1,interactive=True,)
        envelope_ratio_conf = gr.Slider(minimum=0,maximum=1,label="Envelope ratio",value=0.25,interactive=True,)
        consonant_protec_conf = gr.Slider(minimum=0,maximum=0.5,label="Consonant breath protection",value=0.5,interactive=True,)
        button_conf = gr.Button("Convert",variant="primary",)
        output_conf = gr.Audio(type="filepath",label="Output",)

        button_conf.click(
            run,
            inputs=[
                sound_gui,
                pth_file_ui,
                pitch_algo_conf,
                pitch_lvl_conf,
                index_file_ui, # put a bullet through my head
                index_inf_conf,
                respiration_filter_conf,
                envelope_ratio_conf,
                consonant_protec_conf,
            ],
            outputs=[output_conf],
        )
    
    with gr.Tab("Download Model"):
        # markdown
        gr.Markdown(
            "Download the model from the following URL and upload it here. (Hugginface RVC model)"
        )
        model = gr.Textbox(lines=1, label="Model URL")
        download_button = gr.Button("Download Model")
        status = gr.Textbox(lines=1, label="Status", placeholder="Waiting....", interactive=False)
        model_pth = gr.Textbox(lines=1, label="Model pth file", placeholder="Waiting....", interactive=False)
        index_pth = gr.Textbox(lines=1, label="Index pth file", placeholder="Waiting....", interactive=False)
        download_button.click(download_from_url, model, outputs=[status, model_pth, index_pth])
        set_model_button = gr.Button("Set Model")
        #set_model_button.click(

    with gr.Tab("Credits"):
        gr.Markdown(
            """
            Ilaria RVC made by [Ilaria](https://huggingface.co/TheStinger) suport her on [ko-fi](https://ko-fi.com/ilariaowo)
            
            The Inference code is made by [r3gm](https://huggingface.co/r3gm) (his module helped form this space 💖)

            made with ❤️ by [mikus](https://github.com/cappuch) - i hacked it up lel
            """
        )

demo.queue(api_open=False).launch(show_api=False) # idk ilaria if you want or dont want to