Third-Party Software Notices and Attributions

About This Document

This document lists the third-party software components and models that are bundled with or used by PulsarVoice, the desktop application published by PulsarOS Intelligence Inc. (a Canadian federal corporation under the Canada Business Corporations Act, corporation number 17777191, business number 730610631RC0001).

Each listed component is the property of its respective owner or owners and is used by PulsarVoice under the terms of the license that accompanies that component. PulsarVoice's own source code, design, and application are not open-source and remain the property of PulsarOS Intelligence Inc.; nothing in this document grants any rights in the PulsarVoice application itself.

The full and unmodified text of each applicable license is included with the application and is available to every user. Where a component requires that its copyright notice or attribution be reproduced, that notice is reproduced both in this document and in the license materials shipped with the application.

Counsel note: This list, including the component names, the license identifiers, and the attribution lines, is to be confirmed and completed by counsel against the actual versions shipped in each released build of PulsarVoice. License identifiers below are stated as commonly reported and must be verified against the license file actually distributed with each component in the shipped version. Items marked with bracketed placeholders, such as [VERSION] and [COPYRIGHT HOLDERS], are to be filled in by counsel from the shipped artifacts.

Contact for questions about these notices: [legal@pulsaros.ca].

How These Components Are Used in PulsarVoice

PulsarVoice is a free desktop application for macOS, Windows, and Linux that gives terminal AI agents a local voice. Speech synthesis (the agent speaking) and speech-to-text (hearing the user) run entirely on the user's own computer using local models. The components below provide the on-device speech models, the model runtimes that execute them, the audio file handling, the bundled Python interpreter, and a local integrity-check routine. None of these components transmit user audio, transcripts, or text off the device as used in PulsarVoice.

Bundled Components

Kokoro (text-to-speech model) and kokoro-onnx

• What it is used for: Provides one of the on-device text-to-speech voices, allowing the agent to speak its work out loud locally. kokoro-onnx is the runtime glue used to execute the Kokoro model.
• Commonly stated license: Apache License, Version 2.0 (to be confirmed by counsel).
• Attribution / notice: Kokoro and kokoro-onnx, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the Apache License, Version 2.0. Used without modification to the license terms. Full license text included with the application.

OpenAI Whisper model weights, via faster-whisper / CTranslate2

• What it is used for: Provides on-device speech-to-text, transcribing the user's spoken replies locally. faster-whisper and the CTranslate2 engine run the Whisper model efficiently on the user's machine.
• Commonly stated license: MIT License (to be confirmed by counsel; the Whisper model weights and each library are to be verified separately).
• Attribution / notice: Whisper, Copyright (c) [COPYRIGHT HOLDERS]; faster-whisper and CTranslate2, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the MIT License. Full license texts included with the application.

mlx-whisper / MLX (Apple silicon)

• What it is used for: Provides an Apple-silicon-optimized path for on-device speech-to-text on supported Mac hardware, running the Whisper model through the MLX framework.
• Commonly stated license: MIT License (to be confirmed by counsel).
• Attribution / notice: mlx-whisper and MLX, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the MIT License. Full license text included with the application.

Kyutai pocket-tts / Kyutai TTS model

• What it is used for: Provides an additional on-device text-to-speech model option, allowing the agent to speak locally.
• Commonly stated license: The Kyutai TTS model is commonly distributed under the Creative Commons Attribution 4.0 International License (CC-BY-4.0). The accompanying code may carry a separate license; both are to be confirmed by counsel.
• Attribution / notice (required): This component includes the Kyutai TTS model, created by [COPYRIGHT HOLDERS / AUTHORS], licensed under the Creative Commons Attribution 4.0 International License (CC-BY-4.0). The CC-BY-4.0 license requires attribution to the original creators. No endorsement by the creators is implied. The full license text and a link to the license are included with the application. Counsel note: because CC-BY-4.0 requires attribution, this notice must remain visible to users and must be confirmed to satisfy the license's attribution requirements (creator name, license name, and any modifications indicator) for the shipped model version.

onnxruntime

• What it is used for: Runs ONNX-format models on the user's machine, including the text-to-speech path that uses an ONNX model.
• Commonly stated license: MIT License (to be confirmed by counsel).
• Attribution / notice: ONNX Runtime, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the MIT License. Full license text included with the application.

PyTorch (torch)

• What it is used for: Provides a model runtime used by parts of the on-device speech pipeline.
• Commonly stated license: BSD 3-Clause License (to be confirmed by counsel).
• Attribution / notice: PyTorch, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the BSD 3-Clause License. Full license text included with the application.

NumPy

• What it is used for: Provides numerical array operations used in audio and model processing on the user's machine.
• Commonly stated license: BSD 3-Clause License (to be confirmed by counsel).
• Attribution / notice: NumPy, Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the BSD 3-Clause License. Full license text included with the application.

soundfile

• What it is used for: Reads and writes audio files on the user's machine as part of the local voice path.
• Commonly stated license: BSD License (to be confirmed by counsel; the specific BSD variant is to be verified).
• Attribution / notice: soundfile (PySoundFile), Copyright (c) [COPYRIGHT HOLDERS]. Licensed under the BSD License. Full license text included with the application.

python-build-standalone distributions of CPython

• What it is used for: Provides the bundled Python interpreter (CPython) that runs the local speech pipeline, distributed in the python-build-standalone form so the application can run without a separately installed Python.
• Commonly stated license: CPython is distributed under the Python Software Foundation License Agreement (PSF License Agreement). The python-build-standalone distribution may aggregate additional components, each under its own license, all to be confirmed by counsel.
• Attribution / notice: Python and CPython, Copyright (c) [COPYRIGHT HOLDERS, including the Python Software Foundation]. Used under the Python Software Foundation License Agreement. Additional components included in the python-build-standalone distribution are governed by their respective licenses. Full license texts included with the application.

Vendored Ed25519 reference implementation

• What it is used for: A vendored Ed25519 reference implementation used only to verify the application's own integrity signature locally on the user's machine. It does not process user audio, transcripts, or text.
• Commonly stated license / origin: Public domain or reference-implementation origin (to be confirmed by counsel against the actual vendored source and its stated terms).
• Attribution / notice: Includes a vendored Ed25519 reference implementation, originating from [SOURCE / AUTHORS], used under its stated terms ([public domain / reference origin], to be confirmed). Where a copyright notice is required by the source terms, that notice is reproduced in the license materials included with the application.

User-Provided Dependency (Not Bundled)

ffmpeg

• What it is used for: PulsarVoice uses ffmpeg, when present on the user's system, for audio capture and playback.
• Important: ffmpeg is a separate dependency that the user provides and installs on their own system. PulsarVoice invokes ffmpeg if it is available, but does not bundle, include, or redistribute ffmpeg or any of its components. ffmpeg is the property of its respective owners and is licensed under its own terms (commonly the LGPL or GPL family, depending on the build the user installs). Users are responsible for obtaining ffmpeg and for complying with the license that applies to the build they install. PulsarOS Intelligence Inc. makes no representation as to which ffmpeg build a user installs.

General Notices

• Each component above is provided by its owner or owners on the terms of the license that applies to that component. Those licenses, and not this summary, govern the use of each component. In the event of any conflict between this summary and a component's actual license text, the component's license text controls.
• Many of the licenses above (for example, the MIT, BSD, and Apache-2.0 licenses, and the Python Software Foundation License Agreement) include disclaimers of warranty and limitations of liability that run in favor of the component owners. Those disclaimers and limitations apply to the corresponding components as used in PulsarVoice.
• The inclusion of a component in this list does not imply that its owner endorses PulsarVoice or PulsarOS Intelligence Inc.
• The full text of every applicable license is included with the application. If you would like a copy of the license materials, or have a question about an attribution, please contact [legal@pulsaros.ca] or write to PulsarOS Intelligence Inc. at [REGISTERED ADDRESS].

Privacy Note

PulsarVoice runs its voice features entirely on the user's own computer. The components listed above are used locally; the on-device speech path makes no network connection, and no audio, transcript, or text leaves the user's computer through these components. For full details on data handling, see the PulsarVoice Privacy Policy at voice.pulsaros.ca, and direct privacy questions to [privacy@pulsaros.ca].

Governing Law

This document and the use of the components described in it are governed, as between the user and PulsarOS Intelligence Inc., by the laws of the Province of [ONTARIO] and the federal laws of Canada applicable therein, without regard to conflict-of-laws principles. Each third-party component remains subject to the terms of its own license. This governing-law statement does not displace any mandatory consumer or data-protection rights a user may have under applicable law, including the Canadian Personal Information Protection and Electronic Documents Act (PIPEDA), the EU and UK General Data Protection Regulation (GDPR), and the California Consumer Privacy Act as amended by the California Privacy Rights Act (CCPA/CPRA). Because PulsarVoice collects no personal data through the components described here, most such obligations are satisfied by design.

*End of draft. To be confirmed and completed by counsel against the actual shipped versions of PulsarVoice and the components listed above.*