Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import shutil
|
| 3 |
+
import subprocess
|
| 4 |
+
import tempfile
|
| 5 |
+
from pathlib import Path
|
| 6 |
+
|
| 7 |
+
import gradio as gr
|
| 8 |
+
import torchaudio
|
| 9 |
+
import torch
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
# ================================
|
| 13 |
+
# CONFIG
|
| 14 |
+
# ================================
|
| 15 |
+
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
| 16 |
+
MODEL = "htdemucs" # best quality vocal separation
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
# ================================
|
| 20 |
+
# SPLIT FUNCTION
|
| 21 |
+
# ================================
|
| 22 |
+
def split_vocals(audio_file):
|
| 23 |
+
if audio_file is None:
|
| 24 |
+
return None, None, "β Please upload an audio file."
|
| 25 |
+
|
| 26 |
+
work_dir = tempfile.mkdtemp(prefix="demucs_")
|
| 27 |
+
|
| 28 |
+
try:
|
| 29 |
+
print(f"[Demucs] Input: {audio_file}")
|
| 30 |
+
print(f"[Demucs] Device: {DEVICE}")
|
| 31 |
+
|
| 32 |
+
# Run Demucs
|
| 33 |
+
result = subprocess.run(
|
| 34 |
+
[
|
| 35 |
+
"python", "-m", "demucs",
|
| 36 |
+
"--two-stems", "vocals", # only split vocals vs everything else
|
| 37 |
+
"-n", MODEL,
|
| 38 |
+
"-o", work_dir,
|
| 39 |
+
audio_file,
|
| 40 |
+
],
|
| 41 |
+
capture_output=True,
|
| 42 |
+
text=True,
|
| 43 |
+
)
|
| 44 |
+
|
| 45 |
+
if result.returncode != 0:
|
| 46 |
+
error = result.stderr[-1000:] # last 1000 chars to avoid huge dumps
|
| 47 |
+
print(f"[Demucs] Error:\n{error}")
|
| 48 |
+
return None, None, f"β Demucs failed:\n{error}"
|
| 49 |
+
|
| 50 |
+
# Find output files
|
| 51 |
+
track_name = Path(audio_file).stem
|
| 52 |
+
demucs_out = Path(work_dir) / MODEL / track_name
|
| 53 |
+
|
| 54 |
+
vocals_src = demucs_out / "vocals.wav"
|
| 55 |
+
background_src = demucs_out / "no_vocals.wav"
|
| 56 |
+
|
| 57 |
+
if not vocals_src.exists() or not background_src.exists():
|
| 58 |
+
return None, None, "β Demucs ran but output files not found."
|
| 59 |
+
|
| 60 |
+
# Copy to stable paths
|
| 61 |
+
vocals_dst = str(Path(work_dir) / "vocals.wav")
|
| 62 |
+
background_dst = str(Path(work_dir) / "background.wav")
|
| 63 |
+
|
| 64 |
+
shutil.copy2(vocals_src, vocals_dst)
|
| 65 |
+
shutil.copy2(background_src, background_dst)
|
| 66 |
+
|
| 67 |
+
print(f"[Demucs] β
Done. Vocals: {vocals_dst}")
|
| 68 |
+
return vocals_dst, background_dst, "β
Split complete!"
|
| 69 |
+
|
| 70 |
+
except Exception as e:
|
| 71 |
+
return None, None, f"β Exception: {str(e)}"
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
# ================================
|
| 75 |
+
# GRADIO UI
|
| 76 |
+
# ================================
|
| 77 |
+
with gr.Blocks(title="CleanSong AI β Demucs Splitter") as demo:
|
| 78 |
+
|
| 79 |
+
gr.Markdown("# πΈ CleanSong AI β Vocal Splitter")
|
| 80 |
+
gr.Markdown(
|
| 81 |
+
"Upload a song β get the isolated **vocals** and **background** tracks back separately. "
|
| 82 |
+
"Powered by Demucs `htdemucs` model."
|
| 83 |
+
)
|
| 84 |
+
|
| 85 |
+
with gr.Row():
|
| 86 |
+
with gr.Column():
|
| 87 |
+
audio_input = gr.Audio(
|
| 88 |
+
label="Upload Song",
|
| 89 |
+
type="filepath",
|
| 90 |
+
sources=["upload"],
|
| 91 |
+
)
|
| 92 |
+
split_btn = gr.Button("πΈ Split Vocals", variant="primary", size="lg")
|
| 93 |
+
|
| 94 |
+
with gr.Column():
|
| 95 |
+
status = gr.Textbox(label="Status", interactive=False)
|
| 96 |
+
vocals_out = gr.Audio(label="Vocals Only", type="filepath")
|
| 97 |
+
background_out = gr.Audio(label="Background / Instrumental", type="filepath")
|
| 98 |
+
|
| 99 |
+
split_btn.click(
|
| 100 |
+
fn=split_vocals,
|
| 101 |
+
inputs=[audio_input],
|
| 102 |
+
outputs=[vocals_out, background_out, status],
|
| 103 |
+
)
|
| 104 |
+
|
| 105 |
+
gr.Markdown(
|
| 106 |
+
"""
|
| 107 |
+
---
|
| 108 |
+
**Part of the CleanSong AI pipeline.**
|
| 109 |
+
- Upload your song here first to get isolated vocals
|
| 110 |
+
- Feed the vocals into the Whisper Transcriber Space
|
| 111 |
+
- The background track is kept intact for the final remix
|
| 112 |
+
"""
|
| 113 |
+
)
|
| 114 |
+
|
| 115 |
+
if __name__ == "__main__":
|
| 116 |
+
demo.launch()
|