update libs

README.md

@@ -1,8 +1,10 @@
 ---
 title: Pop2Piano Demo
 emoji: 🎹
+python_version: 3.10
+models: ["sweetcocoa/pop2piano"]
 sdk: gradio
-sdk_version: 3.42.0
+sdk_version: 4.39.0
 app_file: app.py
 pinned: true
 ---

app.py

@@ -1,18 +1,15 @@
-import os
 import binascii
-import warnings
+import os
 
 import gradio as gr
 import librosa
 import numpy as np
-import torch
 import pretty_midi
-import pytube as pt
-
-from pytube.exceptions import VideoUnavailable
+import torch
+import yt_dlp
 from transformers import Pop2PianoForConditionalGeneration, Pop2PianoProcessor
 
-from utils import mp3_write, normalize
+from utils import cli_to_api, mp3_write, normalize
 
 yt_video_dir = "./yt_dir"
 outputs_dir = "./midi_wav_outputs"
@@ -26,14 +23,20 @@ composers = model.generation_config.composer_to_feature_token.keys()
 
 
 def get_audio_from_yt_video(yt_link: str):
-    try:
-        yt = pt.YouTube(yt_link)
-        t = yt.streams.filter(only_audio=True)
-        filename = os.path.join(yt_video_dir, binascii.hexlify(os.urandom(8)).decode() + ".mp4")
-        t[0].download(filename=filename)
-    except VideoUnavailable as e:
-        warnings.warn(f"Video Not Found at {yt_link} ({e})")
-        filename = None
+    filename = binascii.hexlify(os.urandom(8)).decode() + ".mp3"
+    filename = os.path.join(yt_video_dir, filename)
+    yt_opt = cli_to_api(
+        [
+            "--extract-audio",
+            "--audio-format",
+            "mp3",
+            "--restrict-filenames",
+            "-o",
+            filename,
+        ]
+    )
+    with yt_dlp.YoutubeDL(yt_opt) as ydl:
+        ydl.download([yt_link])
 
     return filename, filename
 
@@ -82,7 +85,7 @@ block = gr.Blocks()
 with block:
     gr.HTML(
         """
-            <div style="text-align: center; max-width: 700px; margin: 0 auto;">
+            <div style="text-align: center; max-width: 400px; margin: 0 auto;">
               <div
                 style="
                   display: inline-flex;
@@ -131,7 +134,7 @@ with block:
             <div> <h3> <center> Listen to the generated MIDI. </h3> </div>
             """
         )
-        with gr.Row().style(mobile_collapse=False, equal_height=True):
+        with gr.Row(equal_height=True):
             stereo_mix1 = gr.Audio(label="Listen to the Stereo Mix")
             wav_output1 = gr.Audio(label="Listen to the Generated MIDI")
 

pyproject.toml

@@ -0,0 +1,4 @@
+[tool.ruff]
+line-length=100
+select = ["F", "I"]
+ignore = []

requirements.txt

@@ -4,7 +4,7 @@ pretty-midi==0.2.9
 essentia==2.1b6.dev1034
 pyFluidSynth==1.3.0
 transformers
-pytube
-gradio
+yt-dlp>=2024.7.25
+gradio  
 resampy
 pydub

utils.py

@@ -1,5 +1,7 @@
 import numpy as np
 import pydub
+import yt_dlp
+import yt_dlp.options
 
 
 def mp3_write(f: str, sr: int, x: np.ndarray, normalized: bool = False):
@@ -8,14 +10,58 @@ def mp3_write(f: str, sr: int, x: np.ndarray, normalized: bool = False):
         y = np.int16(x * 2**15)
     else:
         y = np.int16(x)
-    song = pydub.AudioSegment(y.tobytes(), frame_rate=sr, sample_width=2, channels=channels)
+    song = pydub.AudioSegment(
+        y.tobytes(), frame_rate=sr, sample_width=2, channels=channels
+    )
     song.export(f, format="mp3", bitrate="256k")
 
 
-def normalize(audio: np.ndarray, min_y: float = -1.0, max_y: float = 1.0, eps: float = 1e-8):
+def normalize(
+    audio: np.ndarray, min_y: float = -1.0, max_y: float = 1.0, eps: float = 1e-8
+):
     max_y -= eps
     min_y += eps
     amax = audio.max()
     amin = audio.min()
     audio = (max_y - min_y) * (audio - amin) / (amax - amin) + min_y
     return audio
+
+
+# yt_dlp script copied from https://github.com/yt-dlp/yt-dlp/blob/28d485714fef88937c82635438afba5db81f9089/devscripts/cli_to_api.py
+create_parser = yt_dlp.options.create_parser
+
+
+def parse_patched_options(opts):
+    patched_parser = create_parser()
+    patched_parser.defaults.update(
+        {
+            "ignoreerrors": False,
+            "retries": 0,
+            "fragment_retries": 0,
+            "extract_flat": False,
+            "concat_playlist": "never",
+        }
+    )
+    yt_dlp.options.create_parser = lambda: patched_parser
+    try:
+        return yt_dlp.parse_options(opts)
+    finally:
+        yt_dlp.options.create_parser = create_parser
+
+
+default_opts = parse_patched_options([]).ydl_opts
+
+
+def cli_to_api(opts, cli_defaults=False):
+    opts = (yt_dlp.parse_options if cli_defaults else parse_patched_options)(
+        opts
+    ).ydl_opts
+
+    diff = {k: v for k, v in opts.items() if default_opts[k] != v}
+    if "postprocessors" in diff:
+        diff["postprocessors"] = [
+            pp
+            for pp in diff["postprocessors"]
+            if pp not in default_opts["postprocessors"]
+        ]
+    return diff