import gradio as gr, glob, os, auditok, zipfile, wave, pytube.exceptions, urllib.error, requests, json
from pytube import YouTube
from moviepy.editor import *
import traceback, yt_dlp
def download_video(url, download_as, use_ytdlp):
if use_ytdlp == True:
try:
ydl_opts = {
'format': f"{download_as}/bestaudio/best",
'postprocessors': [{
'key': 'FFmpegExtractAudio',
'preferredcodec': download_as,
}]
}
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
ydl.download(url)
for i in glob.glob(f"*.{download_as}"):
if os.path.exists(i):
os.rename(i, f"output.{download_as}")
return "Finished downloading! Please proceed to next tab."
except:
raise gr.Error(traceback.format_exc())
else:
try:
yt = YouTube(url)
except pytube.exceptions.RegexMatchError:
raise gr.Error("URL not valid or is empty! Please fix the link or enter one!")
except urllib.error.HTTPError as not_ok:
raise gr.Error(f"Recieved {not_ok}")
except pytube.exceptions.AgeRestrictedError:
raise gr.Error("The video you inputted is age-restricted! Please try another link.")
video = yt.streams.get_highest_resolution()
video.download()
video_path = f"{video.default_filename}"
video_clip = VideoFileClip(video_path)
audio_clip = video_clip.audio
if download_as == "wav":
audio_clip.write_audiofile("output.wav")
elif download_as == "mp3":
audio_clip.write_audiofile("output.mp3")
audio_clip.close()
video_clip.close()
for removalmp4 in glob.glob("*.mp4"):
os.remove(removalmp4)
return "Finished downloading! Please proceed to next tab."
def split_audio_from_yt_video(mindur, maxdur, name_for_split_files, show_amount_of_files_and_file_dur):
if show_amount_of_files_and_file_dur == True:
gr.Warning(f"show_amount_of_files_and_file_dur set to True. This will take longer if your audio file is long.")
if not os.path.exists("output.mp3") and not os.path.exists("output.wav"):
raise gr.Error("Neither output.mp3 or output.wav exist! Did the video download correctly?")
if mindur == maxdur:
raise gr.Error(f"Cannot split mindur={mindur} and maxdur={maxdur}, min and max are the same number.")
elif mindur > maxdur:
raise gr.Error(f"Cannot split mindur={mindur} and maxdur={maxdur}, mindur is higher than maxdur.")
elif name_for_split_files == None:
raise gr.Error("Split files name cannot be empty!")
else:
audio_path = "output.wav" if not os.path.exists("output.mp3") else "output.mp3"
audio_regions = auditok.split(
audio_path,
min_dur=mindur,
max_dur=maxdur,
max_silence=0.3,
energy_threshold=45
)
os.remove(audio_path)
for i, r in enumerate(audio_regions):
filename = r.save(f"{name_for_split_files}-{i+1}.wav")
for f in sorted(glob.glob("*.wav")):
audio_files = glob.glob("*.wav")
zip_file_name = "audio_files.zip"
with zipfile.ZipFile(zip_file_name, "w") as zip_file:
for audio_file in audio_files:
zip_file.write(audio_file, os.path.basename(audio_file))
if show_amount_of_files_and_file_dur == False:
for file2 in glob.glob("*.wav"):
os.remove(file2)
return "Files split successfully!\nCheck below for zipped files.", zip_file_name
elif show_amount_of_files_and_file_dur == True:
largest_file = ("", 0)
total_files = 0
total_length = 0.0
for file_name in glob.glob("*.wav"):
file_path = os.path.join(os.getcwd(), file_name)
if file_path.lower().endswith(".wav"):
try:
with wave.open(file_path, 'r') as audio_file:
frames = audio_file.getnframes()
rate = audio_file.getframerate()
duration = frames / float(rate)
file_size = os.path.getsize(file_path)
if file_size > largest_file[1]:
largest_file = (file_name, file_size)
total_length += duration
total_files += 1
except wave.Error as e:
raise gr.Error(f"Error reading file: {e}")
length_mins = total_length / 60
for file2 in glob.glob("*.wav"):
os.remove(file2)
return f"Files split successfully!\nCheck below for zipped files.\n\n{total_files} files created, {length_mins:.2f} minutes total.", zip_file_name
def split_wav_or_mp3_file(audiofileuploader, mindur2, maxdur2, name_for_split_files2, strict):
if audiofileuploader == None:
raise gr.Error("Audio file cannot be empty!")
if mindur2 == maxdur2:
raise gr.Error(f"Cannot split mindur={mindur} and maxdur={maxdur}, min and max are the same number.")
elif mindur2 > maxdur2:
raise gr.Error(f"Cannot split mindur={mindur} and maxdur={maxdur}, mindur is higher than maxdur.")
elif name_for_split_files2 == None:
raise gr.Error("Split files name cannot be empty!")
else:
audio_path = audiofileuploader
audio_regions = auditok.split(
audio_path,
min_dur=mindur2,
max_dur=maxdur2,
max_silence=0.3,
energy_threshold=45,
strict_min_dur=True if strict == True else False
)
os.remove(audio_path)
for i, r in enumerate(audio_regions):
filename = r.save(f"{name_for_split_files2}-{i+1}.wav")
for f in sorted(glob.glob("*.wav")):
audio_files = glob.glob("*.wav")
zip_file_name2 = "audio_files.zip"
with zipfile.ZipFile(zip_file_name2, "w") as zip_file:
for audio_file in audio_files:
zip_file.write(audio_file, os.path.basename(audio_file))
for file2 in glob.glob("*.wav"):
os.remove(file2)
return f"File split successfully!\nCheck below for zipped files.\nAmount created: {len(audio_files)}", zip_file_name2
def download_video_as_audio_only(yt_video, audio_output_format):
try:
yt = YouTube(yt_video)
except pytube.exceptions.RegexMatchError:
raise gr.Error("URL not valid or is empty! Please fix the link or enter one!")
except urllib.error.HTTPError as not_ok:
raise gr.Error(f"Recieved {not_ok}")
except pytube.exceptions.AgeRestrictedError:
raise gr.Error("The video you inputted is age-restricted! Please try another link.")
video = yt.streams.get_highest_resolution()
video.download()
video_path = f"{video.default_filename}"
video_clip = VideoFileClip(video_path)
audio_clip = video_clip.audio
if audio_output_format == "wav":
audio_clip.write_audiofile("output.wav")
elif audio_output_format == "mp3":
audio_clip.write_audiofile("output.mp3")
audio_clip.close()
video_clip.close()
for mp4remove in glob.glob("*.mp4"):
os.remove(mp4remove)
single_zip_name = "only_audio.zip"
audio_files = glob.glob("*.wav") if audio_output_format == "wav" else glob.glob("*.mp3")
with zipfile.ZipFile(single_zip_name, 'w') as zip_file:
for audio_file in audio_files:
zip_file.write(audio_file, os.path.basename(audio_file))
for outputwavremoval in glob.glob("*.wav"):
if os.path.exists(outputwavremoval):
os.remove(outputwavremoval)
for outputmp3removal in glob.glob("*.mp3"):
if os.path.exists(outputmp3removal):
os.remove(outputmp3removal)
return f"Done! Download the zip file below! This only contains the audio file.\n\nYou have downloaded {yt.title} by {yt.author}.", single_zip_name
def mp4_to_wav_or_mp3(mp4fileuploader, file_format):
if mp4fileuploader == None:
raise gr.Error("Input cannot be empty!")
else:
try:
if file_format == "wav":
videoinput = AudioFileClip(mp4fileuploader)
videoinput.write_audiofile("convertedaudio.wav")
videoinput.close()
elif file_format == "mp3":
videoinput = AudioFileClip(mp4fileuploader)
videoinput.write_audiofile("convertedaudio.mp3")
videoinput.close()
what_to_return = "convertedaudio.wav" if file_format == "wav" else "convertedaudio.mp3"
os.remove(mp4fileuploader)
return "Converted mp4 file successfully!", what_to_return
except:
raise gr.Error(traceback.format_exc())
def remove_audio_file_from_directory():
for r in glob.glob("*.wav"):
os.remove(r)
for w in glob.glob("*.mp3"):
os.remove(w)
return gr.Info("File removed.")
def mvsep_api_request(mvsep_key, audio_file, sep_int):
url = "https://mvsep.com/api/separation/create"
files = {
"audiofile": open(audio_file, 'rb')
}
data = {
"api_token": mvsep_key,
"sep_type": sep_int
}
r = requests.post(url, files=files, data=data)
json_format = r.json()
hash_val = json_format['data']['hash']
return f"Request sent successfully! Your hash is: {hash_val}.\n\nUse the next tab to check the status of your request."
def mvsep_check_request(hash_textbox):
try:
url = "https://mvsep.com/api/separation/get"
params = {
"hash": hash_textbox
}
r = requests.get(url, params=params)
rjson = r.json()
success = rjson['success']
status = rjson['status']
return f"Was successful? {str(success)}.\n Status: {status}."
except requests.exceptions.JSONDecodeError:
return gr.Info("Status not available or request not sent.")
def mvsep_download_separated_audio(hash_textbox):
try:
url = "https://mvsep.com/api/separation/get"
params = {
"hash": hash_textbox
}
r = requests.get(url, params=params)
rjson = r.json()
files = rjson.get('data', {}).get('files', [])
urls = [file['url'] for file in files]
return json.dumps(urls, indent=4)
except requests.exceptions.JSONDecodeError:
return gr.Info("Nothing to download yet. Check back later.")
def mvsep_list_available_models():
return [
"0 - spleeter (vocals, music)",
"1 - spleeter (vocals, drums, bass, other)",
"2 - spleeter (vocals, drums, bass, piano, other)",
"3 - unmix XL (vocals, drums, bass, other)",
"4 - unmix HQ (vocals, drums, bass, other)",
"5 - unmix SD (vocals, drums, bass, other)",
"6 - unmix SE (vocals, music)",
"7 - MDX A (vocals, drums, bass, other)",
"8 - MDX B (vocals, drums, bass, other)",
"9 - UVR HQ (vocals, music)",
"10 - Demucs3 Model A (vocals, drums, bass, other)",
"11 - Demucs3 Model B (vocals, drums, bass, other)",
"12 - MDX-B Karaoke (lead/back vocals)",
"13 - Demucs2 (vocals, drums, bass, other)",
"14 - Zero Shot (Query Based) (LQ)",
"15 - Danna sep (vocals, drums, bass, other)",
"16 - Byte Dance (vocals, drums, bass, other)",
"17 - UVRv5 Demucs (vocals, music)",
"18 - MVSep DNR (music, sfx, speech)",
"19 - MVSep Vocal Model (vocals, music)",
"20 - Demucs4 HT (vocals, drums, bass, other)",
"22 - FoxJoy Reverb Removal (other)",
"23 - MDX B (vocals, instrumental)",
"24 - MVSep Demucs4HT DNR (dialog, sfx, music)",
"25 - MDX23C (vocals, instrumental)",
"26 - Ensemble (vocals, instrumental) [PREMIUM ONLY]",
"27 - Demucs4 Vocals 2023 (vocals, instrumental)",
"28 - Ensemble (vocals, instrumental, bass, drums, other) [PREMIUM ONLY]",
"29 - MVSep Piano (piano, other)",
"30 - Ensemble All-In (vocals, bass, drums, piano, guitar, lead/back vocals, other) [PREMIUM ONLY]",
"31 - MVSep Guitar (guitar, other)",
"33 - Vit Large 23 (vocals, instrum)",
"34 - MVSep Crowd removal (crowd, other)",
"35 - MVSep MelBand Roformer (vocals, instrum)",
"36 - BandIt Plus (speech, music, effects)",
"37 - DrumSep (kick, snare, cymbals, toms)",
"38 - LarsNet (kick, snare, cymbals, toms, hihat)",
"39 - Whisper (extract text from audio)",
"40 - BS Roformer (vocals, instrumental)",
"41 - MVSep Bass (bass, other)"
]
with gr.Blocks(theme='sudeepshouche/minimalist', title="Global Dataset Maker") as app:
gr.HTML(
" Welcome to the GDMGS! (GlobalDatasetMaker Gradio Space)
"
)
gr.Markdown("## Duplicate this space if you want to make your own changes!")
gr.HTML(
"""
"""
)
gr.Markdown(
"This Space will create a dataset for you and use MVSEP to isolate vocals (EXPERIMENTAL), all automatically. **Please be warned that due to not having a GPU on this Space, some steps might take longer to complete.**"
)
gr.HTML(
" This Space's storage is ephemeral, meaning once you reload this space, all audio files will be lost.
"
)
with gr.Tabs():
with gr.TabItem("Download Video"):
with gr.Row():
with gr.Column():
with gr.Row():
url = gr.Textbox(label="URL")
download_as = gr.Radio(["wav", "mp3"], label="Audio format output", value="wav", info="What should the audio format be output as?")
use_ytdlp = gr.Checkbox(False, label="Use yt_dlp instead of pytube?", info="Sometimes Pytube refuses to download a video. If that happens, check this box to download using yt_dlp instead.")
convertion = gr.Button("Download", variant='primary')
convertion.click(
fn=download_video,
inputs=[url, download_as, use_ytdlp],
outputs=gr.Text(label="Output")
)
with gr.TabItem("Split audio files"):
with gr.Row():
with gr.Column():
with gr.Row():
mindur = gr.Number(label="Min duration", minimum=1, maximum=10, value=1)
maxdur = gr.Number(label="Max duration", minimum=1, maximum=10, value=5)
name_for_split_files = gr.Textbox(label="Name for split files")
show_amount_of_files_and_file_dur = gr.Checkbox(False, label="Show total amount of files and duration?")
splitbtn = gr.Button("Split", variant='primary')
splitbtn.click(
split_audio_from_yt_video,
inputs=[mindur, maxdur, name_for_split_files, show_amount_of_files_and_file_dur],
outputs=[gr.Text(label="Output"), gr.File(label="Zipped files")]
)
with gr.TabItem("Misc tools"):
with gr.Tab("File splitter"):
gr.Markdown("If you would rather split a single WAV or mp3 audio file, use this method instead.")
with gr.Row():
with gr.Column():
with gr.Row():
audiofileuploader = gr.File(file_count='single', file_types=[".wav", ".mp3"], label="WAV or mp3 file")
mindur2 = gr.Number(label="Min duration", minimum=1, maximum=10, value=1)
maxdur2 = gr.Number(label="Max duration", minimum=1, maximum=10, value=5)
name_for_split_files2 = gr.Textbox(label="Name for split files")
strict = gr.Checkbox(True, label="Enable strict duration?", info="Use this option if you want to minimize the '(audio_file) is less than 0.76 seconds' warning on Colab. Keep in mind that this only applies for min duration, max is ignored.")
audiofileuploadbtn = gr.Button("Split", variant='primary')
audiofileuploadbtn.click(
split_wav_or_mp3_file,
[audiofileuploader, mindur2, maxdur2, name_for_split_files2, strict],
[gr.Text(label="Output"), gr.File(label="Zipped files")]
)
with gr.Tab("Audio only download"):
gr.Markdown("If you want to download only the audio (to isolate bgm using UVR, etc), use this method, which will only extract audio and not split the audio.")
with gr.Row():
with gr.Column():
with gr.Row():
yt_video = gr.Textbox(label="URL")
audio_output_format = gr.Radio(["wav", "mp3"], value="wav", label="Download audio as:")
commence_download = gr.Button("Download", variant='primary')
commence_download.click(
download_video_as_audio_only,
[yt_video, audio_output_format],
[gr.Text(label="Output"), gr.File(label="Zipped audio file")]
)
with gr.Tab("MP4 to mp3/wav converter"):
gr.Markdown("If you have an mp4 file, you can convert it to mp3 or wav here. Only click the 'Remove file' button when done.")
with gr.Row():
with gr.Column():
with gr.Row():
mp4fileuploader = gr.File(file_count='single', file_types=[".mp4"], label="mp4 file")
file_format = gr.Radio(["wav", "mp3"], value="mp3", label="Convert mp4 file to:")
convert_btn = gr.Button("Convert", variant='primary')
remove_file_btn = gr.Button("Remove file from directory", variant='secondary')
convert_btn.click(
mp4_to_wav_or_mp3,
[mp4fileuploader, file_format],
[gr.Text(label="Output"), gr.File(label="Converted audio file")]
)
remove_file_btn.click(
remove_audio_file_from_directory,
None,
None
)
with gr.Tab("MVSEP"):
gr.Markdown("**VERY EXPERIMENTAL!** Use MVSEP to isolate audio.\n\n**You will be required to input your API key, but it will not be saved ever, I don't use anything saved here for bad intentions, nor would I have access to it regardless.**")
with gr.Tab("Send Request"):
with gr.Row():
with gr.Column():
with gr.Row():
mvsep_key = gr.Textbox(placeholder="Enter your MVSEP API key.", label="API key")
audio_file = gr.File(file_count='single', file_types=[".mp3"], label="Audio file")
sep_int = gr.Number(11, label="Separation type (default is 11).", minimum=0, maximum=41, interactive=True)
send_req = gr.Button("Send request", variant='primary')
get_available_models = gr.Button("Get available models", variant='secondary')
send_req.click(
mvsep_api_request,
[mvsep_key, audio_file, sep_int],
[gr.Text(label="Output")]
)
get_available_models.click(
mvsep_list_available_models,
None,
[gr.Text(label="Models")]
)
with gr.Tab("Get status of request"):
with gr.Row():
with gr.Column():
with gr.Row():
hash_textbox = gr.Textbox(label="Hash")
check_status = gr.Button("Check status", variant='primary')
download = gr.Button("Download separated audio", variant='secondary')
check_status.click(
mvsep_check_request,
[hash_textbox],
[gr.Text(label="Status")]
)
download.click(
mvsep_download_separated_audio,
[hash_textbox],
[gr.Text(label="Link(s)")]
)
with gr.TabItem("Changelog"):
gr.Markdown("v0.99.4 - Added a button to display the available models for MVSEP.")
gr.Markdown("v0.99.3 - Added MVSEP in Misc Tools. This is VERY EXPERIMENTAL and there will be bugs present.")
gr.Markdown("v0.99.2 - Added an mp4 file converter.")
gr.Markdown("v0.99.1 - Removed very old tools (including the 'Upcoming Features' tab) that did not fit with the nature of the program.")
gr.Markdown("v0.99 - Added 'Strict Duration' mode for the file splitter.")
gr.Markdown("v0.98.2 - Added new upcoming features tab.")
app.launch()