Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

polyvoice

CI PyPI License: MIT

Speaker diarization for Python — who spoke when.

Rust-powered speaker diarization that runs on CPU, fits in ~8.4 MB INT8, and requires zero Python runtime overhead. The wheel uses the same hand-written INT8 kernels as the Rust CLI — no ONNX Runtime. Pipeline v2 with VBx clustering and overlap detection.

Install

pip install polyvoice

Requires Python 3.9+.

Quick start

import polyvoice

# Models auto-download on first run (~8.4 MB INT8)
pipeline = polyvoice.Pipeline.balanced()

result = pipeline.run(samples, sample_rate=16000)

print(f"Speakers: {result['num_speakers']}")
for turn in result["turns"]:
    print(f"Speaker {turn['speaker']}: {turn['start']:.1f}s - {turn['end']:.1f}s")

API

  • polyvoice.Pipeline.balanced(models_cache=None, clusterer=None, vbx_plda_dir=None) — balanced accuracy / speed.
  • polyvoice.Pipeline.mobile(models_cache=None, clusterer=None, vbx_plda_dir=None) — smaller, faster model. clusterer is "vbx" (default, matching the CLI) or "ahc". VBx resolves its PLDA params via vbx_plda_dir, then the POLYVOICE_VBX_PLDA_DIR env var, then a registry download.
  • pipeline.run(samples, sample_rate) → dict with num_speakers and turns.
  • pipeline.run_result(samples, sample_rate) → typed DiarizationResult with .to_json() / .to_rttm() / .to_srt() / .to_vtt() / .to_txt() projections.
  • polyvoice.DiarizationResult.from_json(json) — re-hydrate a saved result.

Performance

Pipeline VoxConverse-test DER (collar 0, overlap-scored) Model size
default (v2+VBx, INT8 kernels) 13.3% ~8.4 MB

Full protocol, collar/averaging disclosure, and competitor numbers: docs/BENCHMARKS.md.

See the full repository for Rust / C / CLI APIs, benchmarks, and development docs.

Release wheels are qualified on CPython 3.12 for Linux x86_64/ARM64, macOS ARM64 and Windows x86_64. Other source-build interpreters are not an implied wheel support promise. See the artifact matrix and checks.

Metadata

Release files for polyvoice 1.0.0rc1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Built distributions (wheels)

Table of built distributions (wheels) for polyvoice 1.0.0rc1
File
polyvoice-1.0.0rc1-cp312-cp312-win_amd64.whl CPython 3.12 CPython 3.12 Windows x86-64 Details
polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_x86_64.whl CPython 3.12 CPython 3.12 Linux glibc 2.34+ x86-64 Details
polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_aarch64.whl CPython 3.12 CPython 3.12 Linux glibc 2.34+ ARM64 Details
polyvoice-1.0.0rc1-cp312-cp312-macosx_11_0_arm64.whl CPython 3.12 CPython 3.12 macOS 11.0+ ARM64 Details

Total release size: 10.1 MB

Release files / polyvoice-1.0.0rc1-cp312-cp312-win_amd64.whl

Download URL polyvoice-1.0.0rc1-cp312-cp312-win_amd64.whl
Size 2.4 MB
Tags CPython 3.12 Windows x86-64
SHA-256 checksum
How to use checksums
d5d63cc714e1a3baa524572b741890d2feb5dcca2c882848933e13d797de4b04
BLAKE2b-256 checksum
How to use checksums
2296d630c3aea2fba04d714a59f6b053de693acbf10a4550d34b504f52e1ad44
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 1, 2026.

Transparency log

Release files / polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_x86_64.whl

Download URL polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_x86_64.whl
Size 2.9 MB
Tags CPython 3.12 Linux glibc 2.34+ x86-64
SHA-256 checksum
How to use checksums
7f6dc8f3b7a554a876a0e0f3ff8725f335724584df4f1d060d5f4d0372ff9abc
BLAKE2b-256 checksum
How to use checksums
f7fb821045e11565dde077f646650afbb105242bfee270221f54a17925e762cb
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 1, 2026.

Transparency log

Release files / polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_aarch64.whl

Download URL polyvoice-1.0.0rc1-cp312-cp312-manylinux_2_34_aarch64.whl
Size 2.6 MB
Tags CPython 3.12 Linux glibc 2.34+ ARM64
SHA-256 checksum
How to use checksums
5fcbcfa4a77e5eb7743f99cd7ce6858f46788324b10b32a79076e337f5680c65
BLAKE2b-256 checksum
How to use checksums
0732a6f1aa091dd09d0b9a24377b630384acff9964e529f4f8618836cc48664a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 1, 2026.

Transparency log

Release files / polyvoice-1.0.0rc1-cp312-cp312-macosx_11_0_arm64.whl

Download URL polyvoice-1.0.0rc1-cp312-cp312-macosx_11_0_arm64.whl
Size 2.2 MB
Tags CPython 3.12 macOS 11.0+ ARM64
SHA-256 checksum
How to use checksums
13417f6ef71f838ee1e8a50f2256369b4b1ca098bae2ee17b421330e319ad0d0
BLAKE2b-256 checksum
How to use checksums
2f04c8ae8f3cf462d4ba108ae229cff26593c4a27f7a7813a80b3c0340fbbe26
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 1, 2026.

Transparency log
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page