audio-separation 0.1.2__tar.gz

audio_separation-0.1.2/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 felipe-infantino
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

audio_separation-0.1.2/PKG-INFO

@@ -0,0 +1,92 @@
+Metadata-Version: 2.4
+Name: audio-separation
+Version: 0.1.2
+Summary: Local vocal removal
+License-File: LICENSE
+Author: Felipe Infantino
+Author-email: felipe.infantino@icloud.com
+Requires-Python: >=3.11,<3.15
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Requires-Dist: demucs (>=4.0.1,<5.0.0)
+Requires-Dist: huggingface-hub (>=0.23,<0.24)
+Requires-Dist: librosa (>=0.11.0,<0.12.0)
+Requires-Dist: pydub (>=0.25.1,<0.26.0)
+Requires-Dist: resampy (>=0.4.3,<0.5.0)
+Requires-Dist: torch (>=2.10,<3.0)
+Requires-Dist: torchaudio (>=2.10,<3.0)
+Requires-Dist: tqdm (>=4.67.3,<5.0.0)
+Description-Content-Type: text/markdown
+
+# audio-separation
+
+Local audio source separation. Drums, Bass, Vocal and other. 
+
+## Install
+
+```bash
+pip install audio-separation
+```
+
+## Usage
+
+```bash
+audio-separation "track.mp3" ./outputs/
+audio-separation "track.mp3" ./outputs/ --device cuda
+audio-separation "track.mp3" ./outputs/ --device mps
+```
+
+Output files are saved as `<track>_<stem>.mp3` in the specified directory.
+
+## Arguments
+
+| Argument | Description |
+|---|---|
+| `input_file` | Path to the input audio file |
+| `output_dir` | Directory where the output MP3 will be saved |
+
+## Options
+
+| Flag | Default | Description |
+|---|---|---|
+| `--device` | `cpu` | Inference device: `cpu`, `cuda`, `mps` |
+
+## Requirements
+
+- Python 3.11–3.14
+- FFmpeg (`brew install ffmpeg` / `apt install ffmpeg`)
+
+> Model weights download automatically on first run from Hugging Face (`felipeinfantino/voice-remover`).
+> Cached at `~/.cache/voice-remover/models/`.
+
+## Development
+
+```bash
+git clone https://github.com/felipeinfantino/audio-separation
+cd audio-separation
+poetry install
+poetry run audio-separation "inputs/Alucinados.mp3" "outputs"
+```
+
+### Extending with a new package
+
+```bash
+poetry add [packagename]
+
+# Check the CLI is still working
+poetry run audio-separation "inputs/Alucinados.mp3" "outputs"
+
+# Verify lockfile is clean
+poetry lock
+
+# Bump version
+poetry version patch   # or minor / major
+```
+
+## License
+
+MIT
+

audio_separation-0.1.2/README.md

@@ -0,0 +1,68 @@
+# audio-separation
+
+Local audio source separation. Drums, Bass, Vocal and other. 
+
+## Install
+
+```bash
+pip install audio-separation
+```
+
+## Usage
+
+```bash
+audio-separation "track.mp3" ./outputs/
+audio-separation "track.mp3" ./outputs/ --device cuda
+audio-separation "track.mp3" ./outputs/ --device mps
+```
+
+Output files are saved as `<track>_<stem>.mp3` in the specified directory.
+
+## Arguments
+
+| Argument | Description |
+|---|---|
+| `input_file` | Path to the input audio file |
+| `output_dir` | Directory where the output MP3 will be saved |
+
+## Options
+
+| Flag | Default | Description |
+|---|---|---|
+| `--device` | `cpu` | Inference device: `cpu`, `cuda`, `mps` |
+
+## Requirements
+
+- Python 3.11–3.14
+- FFmpeg (`brew install ffmpeg` / `apt install ffmpeg`)
+
+> Model weights download automatically on first run from Hugging Face (`felipeinfantino/voice-remover`).
+> Cached at `~/.cache/voice-remover/models/`.
+
+## Development
+
+```bash
+git clone https://github.com/felipeinfantino/audio-separation
+cd audio-separation
+poetry install
+poetry run audio-separation "inputs/Alucinados.mp3" "outputs"
+```
+
+### Extending with a new package
+
+```bash
+poetry add [packagename]
+
+# Check the CLI is still working
+poetry run audio-separation "inputs/Alucinados.mp3" "outputs"
+
+# Verify lockfile is clean
+poetry lock
+
+# Bump version
+poetry version patch   # or minor / major
+```
+
+## License
+
+MIT

audio_separation-0.1.2/app/cli.py

@@ -0,0 +1,22 @@
+import argparse
+import sys
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Run vocal removal on an audio file.")
+    parser.add_argument("input_file", help="Path to the input audio file")
+    parser.add_argument("output_dir", help="Directory where the output MP3 will be saved")
+    parser.add_argument(
+        "--device",
+        default="cpu",
+        help="Compute device: cpu | cuda | mps (default: cpu)",
+    )
+    args = parser.parse_args()
+
+    from app.inference import run_inference
+
+    try:
+        run_inference(args.input_file, args.output_dir, device=args.device)
+    except FileNotFoundError as exc:
+        print(f"Error: {exc}", file=sys.stderr)
+        sys.exit(1)

audio_separation-0.1.2/app/inference.py

@@ -0,0 +1,73 @@
+import time
+from pathlib import Path
+
+import numpy as np
+import soundfile as sf
+import torch
+import torchaudio.functional as F
+from demucs.apply import apply_model
+from demucs.pretrained import get_model
+
+DEMUCS_MODEL = "htdemucs"
+
+_model = None
+
+
+def _load_demucs(device: str) -> torch.nn.Module:
+    """Load htdemucs once and cache it for the process lifetime."""
+    global _model
+    if _model is None:
+        print(f"Loading {DEMUCS_MODEL} (downloading once if not cached)...")
+        _model = get_model(DEMUCS_MODEL)
+        _model.eval()
+    return _model.to(device)
+
+
+def run_inference(input_file: str, output_dir: str, device: str = "cpu") -> dict[str, str]:
+    """Separate *input_file* into drums / bass / other / vocals.
+
+    Returns a dict mapping stem name → output file path.
+    """
+    start = time.perf_counter()
+
+    input_path = Path(input_file)
+    if not input_path.exists():
+        raise FileNotFoundError(f"Input file not found: {input_file}")
+
+    # ── 1. Load model ────────────────────────────────────────────────────────
+    model = _load_demucs(device)
+
+    # ── 2. Load audio ────────────────────────────────────────────────────────
+    data, sr = sf.read(str(input_path), always_2d=True)  # [T, C] float64
+    wav = torch.from_numpy(data.T.astype(np.float32))    # [C, T]
+
+    if sr != model.samplerate:
+        wav = F.resample(wav, sr, model.samplerate)
+
+    if wav.shape[0] == 1:
+        wav = wav.repeat(2, 1)                           # mono → stereo
+
+    wav = wav.unsqueeze(0)                              # [1, C, T]  (batch dim)
+
+    # ── 3. Separate ──────────────────────────────────────────────────────────
+    with torch.no_grad():
+        sources = apply_model(model, wav, device=device)  # [1, stems, C, T]
+
+    sources = sources.squeeze(0)                        # [stems, C, T]
+
+    # ── 4. Save one MP3 per stem ─────────────────────────────────────────────
+    basename = input_path.stem
+    out_dir = Path(output_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    output_paths: dict[str, str] = {}
+    for stem_name, stem_wav in zip(model.sources, sources):
+        out_path = out_dir / f"{basename}_{stem_name}.mp3"
+        audio_np = stem_wav.cpu().numpy().T          # [T, C]
+        sf.write(str(out_path), audio_np, model.samplerate, format="MP3")
+        output_paths[stem_name] = str(out_path)
+        print(f"  {stem_name:8s} → {out_path}")
+
+    elapsed = time.perf_counter() - start
+    print(f"Done in {elapsed:.2f}s")
+    return output_paths

audio_separation-0.1.2/pyproject.toml

@@ -0,0 +1,53 @@
+[tool.poetry]
+name = "audio-separation"
+version = "0.1.2"
+description = "Local vocal removal"
+authors = ["Felipe Infantino <felipe.infantino@icloud.com>"]
+readme = "README.md"
+packages = [{ include = "app" }]
+
+[tool.poetry.dependencies]
+python  = ">=3.11,<3.15"
+pydub   = "^0.25.1"
+librosa = "^0.11.0"
+resampy = "^0.4.3"
+
+# torch & torchaudio are pulled in transitively by demucs.
+# No explicit source: PyPI ships macOS (arm64 + x86_64) wheels natively.
+# On Linux/Windows you can pin source = "pytorch-cpu" for the smaller CPU build.
+torch      = "^2.10"
+torchaudio = "^2.10"
+
+huggingface-hub = "^0.23"
+
+# ---------------------------------------------------------------------------
+# Sources
+# ---------------------------------------------------------------------------
+# Default PyPI
+tqdm = "^4.67.3"
+demucs = "^4.0.1"
+[[tool.poetry.source]]
+name     = "pypi"
+priority = "primary"
+
+# PyTorch CPU-only wheels (smaller, works everywhere).
+# Swap the URL for a CUDA build if you need GPU locally:
+#   https://download.pytorch.org/whl/cu121   (CUDA 12.1)
+#   https://download.pytorch.org/whl/cu124   (CUDA 12.4)
+[[tool.poetry.source]]
+name     = "pytorch-cpu"
+url      = "https://download.pytorch.org/whl/cpu"
+priority = "explicit"
+
+# ---------------------------------------------------------------------------
+# Scripts
+# ---------------------------------------------------------------------------
+[tool.poetry.scripts]
+audio-separation = "app.cli:main"
+
+# ---------------------------------------------------------------------------
+# Build system
+# ---------------------------------------------------------------------------
+[build-system]
+requires      = ["poetry-core>=2.0.0"]
+build-backend = "poetry.core.masonry.api"