juancopi81 commited on
Commit
fa0462c
1 Parent(s): 7c627a2

Allow ismirmodel

Browse files
Files changed (1) hide show
  1. app.py +5 -6
app.py CHANGED
@@ -53,7 +53,7 @@ def get_audio(url):
53
  def populate_metadata(link):
54
  yt = YouTube(link)
55
  audio = get_audio(link)
56
- return yt.thumbnail_url, yt.title, audio, audio
57
 
58
  def inference(yt_audio_path):
59
 
@@ -67,7 +67,7 @@ def inference(yt_audio_path):
67
  note_seq.sequence_proto_to_midi_file(est_ns, "./transcribed.mid")
68
 
69
  synth = note_seq.midi_synth.fluidsynth
70
- array_of_floats = synth(est_ns, sample_rate=SAMPLE_RATE)
71
  int16_data = note_seq.audio_io.float_samples_to_int16(array_of_floats)
72
  # piano_roll = create_image_from_note_sequence(note_sequence)
73
 
@@ -95,7 +95,7 @@ with demo:
95
  The mt3 model transcribes multiple simultaneous instruments, but without velocities.
96
  """
97
  model = gr.Radio(
98
- ["mt3",],
99
  label=model_label,
100
  value="mt3"
101
  )
@@ -107,9 +107,8 @@ with demo:
107
  img = gr.Image(label="Thumbnail")
108
  with gr.Row():
109
  yt_audio = gr.Audio()
110
- yt_audio_path = gr.Textbox(visible=False)
111
 
112
- link.change(fn=populate_metadata, inputs=link, outputs=[img, title, yt_audio, yt_audio_path])
113
 
114
  with gr.Row():
115
  btn = gr.Button("Transcribe music")
@@ -119,7 +118,7 @@ with demo:
119
  midi_audio = gr.Audio()
120
 
121
  btn.click(inference,
122
- inputs=yt_audio_path,
123
  outputs=[midi_file, midi_audio])
124
 
125
  gr.Markdown(article)
 
53
  def populate_metadata(link):
54
  yt = YouTube(link)
55
  audio = get_audio(link)
56
+ return yt.thumbnail_url, yt.title, audio
57
 
58
  def inference(yt_audio_path):
59
 
 
67
  note_seq.sequence_proto_to_midi_file(est_ns, "./transcribed.mid")
68
 
69
  synth = note_seq.midi_synth.fluidsynth
70
+ array_of_floats = synth(est_ns, sample_rate=SAMPLE_RATE, sf2_path=SF2_PATH)
71
  int16_data = note_seq.audio_io.float_samples_to_int16(array_of_floats)
72
  # piano_roll = create_image_from_note_sequence(note_sequence)
73
 
 
95
  The mt3 model transcribes multiple simultaneous instruments, but without velocities.
96
  """
97
  model = gr.Radio(
98
+ ["mt3", "ismir2021"],
99
  label=model_label,
100
  value="mt3"
101
  )
 
107
  img = gr.Image(label="Thumbnail")
108
  with gr.Row():
109
  yt_audio = gr.Audio()
 
110
 
111
+ link.change(fn=populate_metadata, inputs=link, outputs=[img, title, yt_audio])
112
 
113
  with gr.Row():
114
  btn = gr.Button("Transcribe music")
 
118
  midi_audio = gr.Audio()
119
 
120
  btn.click(inference,
121
+ inputs="final_audio.wav",
122
  outputs=[midi_file, midi_audio])
123
 
124
  gr.Markdown(article)