python-vocos

LOW
maintainer Smoolak 0 votes scanned 2026-11-01 20:01:18.389276
View on AUR
Why flagged

Uploaded within the last 14 days with 2 or fewer community votes — little peer review so far.

Triggered rules

Low Few votes, recently uploaded zero_votes_recent

Uploaded within the last 14 days with 2 or fewer community votes — little peer review so far.

PKGBUILD

1# Maintainer: Smoolak <smoolak@gmail.com>
2pkgname=python-vocos
3pkgver=0.1.0
4pkgrel=1
5pkgdesc='Fourier-based neural vocoder for high-quality audio synthesis'
6arch=(any)
7url='https://github.com/charactr-platform/vocos'
8license=(MIT)
9depends=(python python-pytorch python-torchaudio python-numpy python-scipy
10 python-einops python-yaml python-huggingface-hub python-encodec)
11# The public EnCodec VCS provider requires git but does not declare it.
12makedepends=(git python-build python-installer python-setuptools python-wheel)
13optdepends=('python-pytorch-lightning: model training'
14 'python-jsonargparse: training command-line configuration'
15 'python-transformers: training evaluation models'
16 'python-matplotlib: training visualizations')
17_commit=11b4628853c9e79dbc4577b01ecb02cfbc78255d
18source=("vocos-$pkgver.tar.gz::$url/archive/$_commit.tar.gz")
19sha256sums=('9ae624fb3a3d963273b4008eda54d530e6820b5a34d31495d7caa937e7e6d66b')
20
21build() {
22 cd "vocos-$_commit"
23 python -m build --wheel --no-isolation
24}
25
26check() {
27 cd "vocos-$_commit"
28 # This release has no automated suite. Exercise its documented mel-to-audio
29 # inference path offline. Pretrained copy synthesis is checked after install.
30 python -m installer --destdir="$srcdir/check" dist/*.whl
31 local site_packages=("$srcdir"/check/usr/lib/python*/site-packages)
32 cd "$srcdir"
33 PYTHONPATH="${site_packages[0]}" python - <<'PYTEST'
34import torch
35from vocos import Vocos
36from vocos.feature_extractors import MelSpectrogramFeatures
37from vocos.heads import ISTFTHead
38from vocos.models import VocosBackbone
39
40torch.set_num_threads(2)
41model = Vocos(MelSpectrogramFeatures(), VocosBackbone(100, 32, 64, 2),
42 ISTFTHead(32, 1024, 256)).eval()
43audio = torch.sin(torch.arange(24000) * (2 * torch.pi * 440 / 24000))[None] * 0.1
44features = model.feature_extractor(audio)
45reconstructed = model(audio)
46torch.testing.assert_close(reconstructed, model.decode(features))
47assert reconstructed.shape[0] == 1 and 23000 <= reconstructed.shape[1] <= 25000
48assert torch.isfinite(reconstructed).all()
49print('Built-wheel mel extraction and Fourier waveform reconstruction: PASS')
50PYTEST
51}
52
53package() {
54 cd "vocos-$_commit"
55 python -m installer --destdir="$pkgdir" dist/*.whl
56 install -Dm644 LICENSE "$pkgdir/usr/share/licenses/$pkgname/LICENSE"
57}
58

Scan history

Scanned at (UTC)SeverityRules
2026-11-01 20:01:18 Low 1

Report a package

Reports go to the AURWatch maintainer (one person) and are read by hand. No login required.

0 / 4000
Your suggestion