Whisperstream 1.6.x license agreement

WHISPERSTREAM END USER LICENSE AGREEMENT
========================================

Version 1.8.1

Effective 2026-10-26

This End User License Agreement (the "Agreement") is a binding legal
agreement between you (the "Licensee," "you," or "your") and the
Licensor identified in Section 1 below. It governs the private
edition of Whisperstream (the version 1 line). It does not govern
the separate open-source edition of Whisperstream, which has its own
license terms (see Sections 3(e) and 10(b)). It becomes binding only
as described in the Acceptance paragraph below and, for an updated
Agreement, Section 20. If you do not agree to these terms, do not
install or use the Software.

Acceptance. This Agreement becomes binding on you when you (a) click
"I Agree" (or a similarly captioned button) presented by the
installer, by the Software on first run, by the Software when these
terms are amended, or by the Software when it asks you to review and
re-acknowledge the then-current terms, (b) check a box indicating
acceptance during initial setup of the Software, or (c) commence use
of the Software after such acceptance. The Software records the
version and content hash of this Agreement, the date and time of
acceptance, the installed Software version, and the acceptance event
type in an append-only local ledger at
%LOCALAPPDATA%\Whisperstream\acceptance.json. You may review this
record at any time. If you did not have an opportunity to review and
accept this Agreement before the Software began running on your
Device, you may obtain the current version at
https://whisperstream.io/eula/private and exercise the
no-questions-asked refund right described in Section 8 within thirty
(30) days of installation, treating the installation date as the
purchase date.

1. PARTIES
==========

The "Licensor" is Lanreal Technologies Inc., a corporation
incorporated under the Business Corporations Act (Ontario), with its
registered office at 18 King Street East, Suite 1400, Toronto, ON
M5C 1C4, Canada ("Lanreal," "we," "us," or "our").

The "Licensee" is the individual or legal entity that downloads,
installs, purchases, or otherwise uses the Software. If you are
entering into this Agreement on behalf of an entity, you represent
that you have authority to bind that entity to this Agreement, and
references to "you" in this Agreement include both you personally
and that entity.

2. DEFINITIONS
==============

In this Agreement:

- "Software" means the Whisperstream desktop application for
  Microsoft Windows, including the Tauri-based desktop shell, the
  bundled Python speech-recognition backend, all bundled binaries,
  scripts, configuration files, and Documentation, together with any
  updates and patches within the same major version line that
  Lanreal makes available under this Agreement. Any major new
  version of the Software (for example, 2.0) that Lanreal elects to
  release may be offered as a separate product under a separate
  license or purchase, as described in Section 23. The open-source
  edition of Whisperstream described in Sections 3(e) and 10(b) is
  not part of the Software.
- "License Key" means the alphanumeric activation credential issued
  to you by Lanreal (or its merchant of record) upon purchase of a
  Pro License, used to activate Pro features of earlier versions of
  the Software. This version of the Software does not use License
  Keys.
- "Device" means a single physical or virtual computer running a
  supported version of Microsoft Windows on which an instance of the
  Software is installed.
- "Pro License" means a paid license for earlier versions of the
  Software, purchased for a License Key from Lanreal or its merchant
  of record. Lanreal does not currently offer Pro Licenses for sale.
  Your license under Section 3 does not depend on whether you bought
  one.
- "Documentation" means the user-facing documentation for the
  Software that Lanreal makes available, including the README,
  in-app help text, and any user guides published at
  whisperstream.io.
- "Third-Party Components" means the open-source and third-party
  libraries, models, and binaries incorporated into or distributed
  with the Software, as listed in the file THIRD_PARTY_LICENSES.txt
  bundled with the Software.

3. GRANT OF LICENSE
===================

Subject to your compliance with this Agreement, Lanreal grants you a
non-exclusive, non-transferable, and non-sublicensable license to
install and use the Software, with all of its features, for personal
and commercial purposes on Devices that you own or control, as
follows.

(a) Every installation. Every installation of this version of the
Software receives the same license under this Section 3, whether or
not you bought a Pro License for an earlier version. Lanreal charges
no fee for it.

(b) No trial or activation. This license does not depend on a trial,
a License Key, an activation, or a purchase. Live dictation is not
limited by a trial period or a per-session minute cap, and this
version does not start, check, or extend a trial.

(c) Earlier license records. This version does not activate,
validate, deactivate, or transfer License Keys and does not contact
any licensing service. It reads, but never creates or changes, a
license and trial record that an earlier version may have left on
your Device, only to choose the wording of an in-app notice, as
described in Section 4. Whether that record exists, can be read, or
shows a purchase does not affect your license.

(d) No promise of continued availability. The license in this
Section 3 has no fixed contractual expiration. That phrase describes
the duration of the license right. It does not promise that
downloads, updates, support, compatibility, any feature, or the
Software's operation will remain available indefinitely, as
explained in Section 23.

(e) Open-source edition. Lanreal may make a separate open-source
edition of Whisperstream available under its own license terms. That
edition is not governed by this Agreement. This version does not
download or install it; Section 10(b) describes the link to the
Whisperstream website. If you install that edition, its installer
may replace this Software on your Device. The availability,
features, updates, and terms of the open-source edition are not
promised by this Agreement.

Section 8 applies only to Pro License purchases. Every other term of
this Agreement applies to you whether or not you bought a Pro
License.

The license granted in this Section 3 takes effect upon your
acceptance of this Agreement.

Your license is contingent on your continuing compliance with this
Agreement. Lanreal may suspend or terminate your license immediately
if your use materially violates applicable law or poses a material
threat to the security or integrity of the Software or its related
services, or if another material breach is, by its nature, incapable
of cure. A refund, reversal, or chargeback of a Pro License
purchase, including a goodwill refund described in Section 8, does
not suspend or terminate your license. For any other material
breach, Lanreal will provide written notice describing the breach.
If you do not cure it within thirty (30) days after the date of that
notice, your license terminates automatically. Lanreal may take
reasonable interim measures to prevent unauthorized or harmful use
while a cure period is pending. These rights are without prejudice
to any other remedy available to Lanreal and remain subject to
rights that applicable law does not permit Lanreal to exclude.

4. LICENSE KEYS AND EARLIER RECORDS
===================================

(a) No license operations. This version of the Software does not
activate, validate, deactivate, or transfer License Keys and does
not contact Lanreal's licensing provider. You do not need a License
Key to use any feature of this version.

(b) Earlier records. An earlier version may have left a protected
license and trial record for your Windows account on your Device.
This version only reads that record, on your Device, to choose the
wording of an in-app notice about the open-source edition. It never
creates, rewrites, or deletes the record. Removing the record,
reinstalling Windows, changing the Device's name or your Windows
account name, or using another Device may change that notice's
wording, but it does not change your license under Section 3.

5. PERMITTED USE
================

You may use the Software for any lawful personal or commercial
purpose, including but not limited to:

- personal use (note-taking, journaling, accessibility, hobby
  projects);
- paid professional work (writing, software development, customer
  support, content creation);
- dictation in regulated environments, subject to your independent
  compliance with applicable laws and the disclaimers in Section 14;
- use within your own business or place of employment.

The license granted in Section 3 covers your own use of the
Software. It does not authorize you to provide the Software, in
whole or in part, as a service to third parties, nor to make the
Software available to multiple end users from a single installation.

You may choose the installer's all-users option to make the
application files available to multiple Windows accounts on one
Device. That option changes only the technical installation scope.
It does not remove the requirement that each person who uses the
Software accept this Agreement.

6. PROHIBITED USE
=================

You may not, and you may not permit any third party to:

(a) redistribute, sell, rent, lease, lend, sublicense, or otherwise
transfer the Software, the installer for the Software, or any
License Key, except as expressly permitted by this Agreement;

(b) reverse engineer, decompile, disassemble, or otherwise attempt
to derive the source code, underlying ideas, algorithms, file
formats, or non-public protocols of the Software, except, and only
to the extent that, such activity is (i) expressly permitted by
mandatory provisions of applicable law (including any rights of
interoperability that cannot be lawfully waived by contract), or
(ii) necessary to exercise rights granted to you under the licenses
applicable to the Third-Party Components listed in
THIRD_PARTY_LICENSES.txt (including, without limitation, the right
to relink the Software with a modified version of an LGPL-licensed
library);

(c) remove, obscure, alter, or fail to reproduce any copyright,
trademark, patent, attribution, or other proprietary notices
contained in or on the Software, the Documentation, or any
Third-Party Components;

(d) re-sell, rebrand, white-label, or sublicense the Software or any
component of the Software;

(e) use the Software, in whole or in part, to design, develop,
train, test, benchmark, or improve a product or service that
competes directly with the Software;

(f) use the Software in any manner that violates applicable law or
the rights of any third party.

7. AI / ML TRAINING
===================

The Software's ordinary transcription and history pipeline is
designed to operate locally. Except for user-directed transmissions
expressly described in Sections 9 and 12, the Software does not
automatically transmit your audio recordings or transcribed text to
Lanreal. Lanreal does not receive or use audio or transcribed text
through the ordinary local transcription and history pipeline,
including for training, fine-tuning, evaluation, or improvement of
an artificial intelligence or machine-learning model. Text is
delivered at your direction to the focused or pinned destination
application, and optional Agent Speak interactions are exchanged
locally with Claude Code or Codex as described in Section 9. The
Software keeps a local history of your dictations (your transcribed
text and, if you enable it, your audio) stored encrypted at rest on
your own Device, as described in Section 9 of this Agreement and in
Sections 3.4 and 8 of the Whisperstream Privacy Policy.

If you enable the optional cloud post-processing feature described
in Section 12, transcribed text is transmitted to the third-party
provider you selected for the purpose of post-processing only.
Lanreal does not receive, access, or use that text for AI or
machine-learning training purposes. Whether the third-party provider
uses your submitted content for its own model training is governed
by the provider's terms of service, not by this Agreement. See
Section 12(d).

This Section describes the Software governed by this Agreement. A
future feature that transmits audio or transcribed text for model
training would require the material-amendment process in Section 20
and any notice or consent required by applicable law.

8. REFUND POLICY
================

Lanreal offers a 30-day no-questions-asked refund on the license fee
paid at checkout for a Pro License. To request a refund, send an
email to support@whisperstream.io from the email address you used at
checkout within thirty (30) days of your original purchase date.
Approved refunds are processed by the merchant of record and may
take several business days to appear on your original payment
method. The contractual refund described in this Section 8 does not
limit any non-waivable statutory cancellation or cooling-off rights
you may have under the consumer protection laws of your jurisdiction
(see Section 17).

Abuse carve-out. Lanreal reserves the right, acting reasonably and
in good faith, to refuse a refund request where Lanreal reasonably
determines that the refund request constitutes abuse of this refund
policy, including but not limited to: repeated refund requests
across multiple License Keys, accounts, payment methods, or email
addresses associated with the same person; refund requests submitted
shortly after extracting substantial use of the Software beyond what
a reasonable consumer evaluating the Software would require; or
refund requests that form part of a pattern of fraudulent or
chargeback-related activity. Refunds refused under this carve-out
are without prejudice to any non-waivable statutory refund rights
you may have under Section 17.

Goodwill refunds. Lanreal has chosen to refund Pro License purchases
as a voluntary goodwill gesture. Lanreal decides, at its sole
discretion, whether to issue a goodwill refund for any purchase, and
its amount, timing, and method. This paragraph does not create an
obligation for Lanreal to issue a goodwill refund or give you a
right to receive one, and Lanreal's stated intention to issue
goodwill refunds is not a promise to you. Lanreal may issue a
goodwill refund through the merchant of record or, where that is not
possible, contact you at the email address used for the purchase to
arrange another method. You may tell Lanreal that you do not want a
goodwill refund. Your license under Section 3 does not depend on
whether you receive, decline, or are not offered a goodwill refund.
This paragraph does not limit the 30-day refund described above, the
refund right in the Acceptance paragraph, or any non-waivable
statutory rights you may have under Section 17.

9. NETWORK COMMUNICATIONS / PRIVACY
===================================

Network Communications. Whisperstream is designed to operate
locally. In its default configuration the Software makes only the
network connections described in paragraphs (b) and (c) below.
Additional user-initiated flows are described separately: paragraph
(d) covers voluntary feedback submission, paragraph (e) covers the
optional voice feature's model download, paragraph (f) covers the
optional on-device AI cleanup model, paragraph (k) covers the
curated cloud-model list, Section 10(b) covers the link to the
open-source edition's download page, and Section 12 plus Section 6
of the Whisperstream Privacy Policy cover the optional cloud
post-processing feature (Bring Your Own Key), imported audio, Agent
Speak, and the external Featurebase board. None of those
user-initiated flows occurs unless you explicitly take action.

(a) No license operations. This version makes no license activation,
validation, or deactivation request and does not contact Polar
Software Inc. It only reads, on your Device, the earlier license and
trial record described in Section 4(b) and Privacy Policy Section
3.9.

(b) Update checks. The Software checks for signed updates to this
version line at https://releases.whisperstream.io/latest.json,
hosted on Cloudflare, Inc.'s infrastructure. It does not check the
open-source edition's update channel (Section 10(b)). These requests
are standard HTTPS GET requests; no personal information, device
identifier, version string, or other identifying data is placed in
the request URL, headers, or body beyond the standard User-Agent and
Accept headers automatically added by the underlying HTTP library.
Update checks require acceptance of the current Agreement and
completed setup, and they do not run while Lockdown Mode is on,
whether you turned it on or your organization manages it. Update
manifests and installers are cryptographically signed; the Software
verifies signatures locally before installing any update. Section 10
describes the schedule and your controls.

(c) Speech-recognition model downloads. During setup, or after a
language or engine selection, the Software downloads the required
speech recognition model if it is not already installed. It requests
pinned model files from Hugging Face and its file-delivery
infrastructure. If the Hugging Face source is unavailable for
Parakeet, Qwen, or Cohere Transcribe, the Software may instead
request the matching pinned sherpa-onnx release archive from GitHub
and its release-asset service. These services receive the requested
file path, your IP address, and the standard HTTP headers, including
a User-Agent, that accompany an ordinary download request. The
Software does not attach a Hugging Face token, Lanreal account
identifier, license key, device identifier, audio, transcript, or
provider credential to these downloads. Each file or archive is
checked against its expected size and SHA-256 checksum before it is
used. Existing local models can be used without downloading them
again. If you manually select Cohere Transcribe, the requested URL
identifies the Cohere model, but your selected transcription
language remains in local settings and is not included in the model
request. Cohere Transcribe requires you to select one of its
supported languages and runs entirely on your Device. The Software
does not send your audio or transcribed text to Cohere, Hugging
Face, GitHub, Lanreal, or Cloudflare for speech recognition.

(d) Feedback submissions (voluntary only). If you choose to send
private feedback from the Software's About page, the Software
transmits the feedback text and the email address you entered,
together with the application version, to Lanreal's feedback
database hosted on Supabase, Inc.'s infrastructure at supabase.co.
This flow does not occur unless you choose to send the feedback.
This version does not ask for or submit star ratings. For full
details see Section 3.5 of the Whisperstream Privacy Policy.

(e) Voice model downloads (optional voice feature only). If you
enable the optional voice feature described in Section 27, the
Software downloads the Voice Model or Voice Models needed for your
selected or automatically resolved engine from pinned Hugging Face
repositories and Hugging Face's file-delivery infrastructure. The
Kokoro Voice Model archive is published in a Hugging Face repository
operated by Lanreal. The request includes only the standard HTTP
headers automatically added by the underlying download library,
including a User-Agent, and your IP address is visible to the
applicable host as part of the network connection. The Software does
not put your spoken text, generated audio, license key, account
information, or device identifier in the request. Each downloaded
file or archive is verified against a cryptographic checksum built
into the Software before the model is installed. This flow does not
occur unless you enable the voice feature. For full details see
Section 3.7 of the Whisperstream Privacy Policy.

(f) On-device AI cleanup model download (optional). If you select
the on-device AI cleanup option and its Gemma model is not already
installed, the Software downloads the pinned model file from Hugging
Face and its file-delivery infrastructure. The request includes the
same ordinary connection metadata described above, including your IP
address and standard HTTP headers, but does not include transcript
text, a license key, account information, or a device identifier.
The downloaded model is verified locally before installation.

(g) Imported audio. If you import an audio file, the Software reads,
decodes, and transcribes it locally. The audio file is not uploaded.
If you have enabled Section 12 cloud post-processing, the resulting
transcript may be sent to your selected provider under that Section.

(h) Agent Speak, Claude Code, and Codex. If you enable Agent Speak,
the Software's local plugins and loopback backend exchange Claude
Code or Codex responses, permission prompts, replies, limited
session identity, and, for Claude Code, structured questions, so the
local HUD can display, speak, or answer them. The Software does not
automatically send this interaction content to Lanreal. Active
interaction content is kept in bounded process memory; hook
payloads, content-derived fingerprints, session identifiers, paths,
and interaction content are excluded from diagnostic logs and
support exports. A reply you dictate through Agent Speak is stored
as an agent_reply history entry when transcript history is enabled
and follows its encrypted retention settings; other coding-agent
interaction content is not written to transcript history. Agent
Speak does not control the coding agent's own connection to
Anthropic, OpenAI, or another provider you chose; that separate
connection is governed by your coding-agent and provider terms.
While Lockdown Mode is on, the Software disables Agent Speak and
clears its active state, but it does not close Claude Code or Codex
or control their own network connections.

(i) External feature-request board. The Software may open Lanreal's
Featurebase board in your browser when you choose the
feature-request control. No transcript or other application content
is sent automatically. Information you choose to submit is processed
under Section 3.12 and Section 5 of the Whisperstream Privacy Policy
and under Featurebase's terms and privacy policy.

(j) No trial checks. This version has no trial and makes no
trial-extension policy request. The trial_policy_checks_enabled
setting, if present in your configuration file, has no effect.

(k) Curated cloud-model list. When you open the Software's
Enhancement page while a cloud provider is selected for the optional
cloud post-processing feature (a cloud provider is selected by
default, even before you enable or configure that feature), or when
you choose a cloud provider, the Software may request Lanreal's
curated list of supported cloud models from
https://models.whisperstream.io/cloud-models.json, hosted on
Cloudflare, Inc.'s infrastructure, and keep the last valid copy
locally. The request carries only ordinary connection data and
standard HTTP headers. It does not include transcript text, a
provider credential, a license key, or a device identifier. If the
request fails, the Software uses its last valid copy or the list
included with the Software. This request is not made for cloud
providers while Lockdown Mode is on.

Model-host roles and international transfers. Hugging Face and
GitHub receive model-download requests directly from your Device and
determine their own purposes for the connection data they receive,
as described in Sections 5 and 7 of the Whisperstream Privacy
Policy. Lanreal does not engage them to process that data on its
behalf. Cloudflare serves the update and cloud-model-list requests
described above. This version does not request model files or a
model-availability manifest from Lanreal's model endpoint at
models.whisperstream.io; its only request to that endpoint is the
cloud-model list in paragraph (k).

What is NEVER sent (default configuration). In the default
configuration of the Software (without enabling the optional cloud
post-processing feature of Section 12 and without submitting
feedback under (d) above), the Software does not send your audio
recordings, your microphone input, your typing, your keystrokes,
your clipboard contents, or your transcribed text to any server, at
any time, for any purpose. The local backend process binds
exclusively to the loopback interface (127.0.0.1) and exposes no
telemetry endpoints. The Software contains no analytics, error
reporting, crash reporting, or usage tracking service.

Transcribed text is local by default. In the default configuration
of the Software, transcribed text is typed into the focused window
and is not sent to any server. The Software keeps a local, on-device
history of your dictations (your transcribed text and, if you enable
it, your audio), encrypted at rest and stored only on your Device;
that history is never transmitted off your Device and is described
further in Section 7 of this Agreement and in Sections 3.4 and 8 of
the Whisperstream Privacy Policy. If you explicitly enable the
optional cloud post-processing feature described in Section 12 of
this Agreement and Section 6 of the Whisperstream Privacy Policy,
transcribed text longer than three words is transmitted only to the
specific third-party provider you selected, only over HTTPS, and
only for the duration of each post-processing request. If you turn
on Lockdown Mode (Section 14), the optional cloud post-processing
feature is disabled and AI cleanup runs only on the on-device model,
so transcribed text does not leave your Device. This Section 9
describes the network communications that the Software performs on
your behalf, and must be read together with Section 12, which
governs your use of the optional cloud post-processing feature and
your relationship with the provider. No transcribed text is sent to
Lanreal, Polar, Cloudflare, Supabase, Hugging Face, or GitHub as
part of Whisperstream's transcription or cloud-cleanup pipeline.
Text you deliberately deliver to another application, submit through
a feedback destination, or exchange with Claude Code or Codex is
handled by that destination under its own terms and privacy
practices.

The Whisperstream Privacy Policy is available at
https://whisperstream.io/privacy/private and is incorporated by
reference.

10. UPDATES AND THE OPEN-SOURCE EDITION
=======================================

(a) Updates to this version line. The Software includes an update
mechanism that checks the manifest at
https://releases.whisperstream.io/latest.json to determine whether a
newer release in this version line is available. The Software
attempts an automatic check when it starts, when its window gains
focus, and on an hourly timer, but it sends no more than one
automatic check every 24 hours, backing off to no more than one
every seven days after repeated failures. You can also check
manually from the About page. Checks require acceptance of the
current Agreement and completed setup and do not run while Lockdown
Mode is on. If you decline an amendment to this Agreement that the
Software presents, the Software may suspend automatic update checks
for a limited period, as described in Section 20. When an update is
available, the Software shows you its release notes and downloads
and installs it only when you choose to update. Update manifests and
installers are cryptographically signed using the minisign signing
scheme, and the Software verifies each signature locally against an
embedded public key before any update is applied. Updates that fail
signature verification are rejected.

(b) The open-source edition. This version does not check the
open-source edition's update channel, show or record acceptance of
its terms, or download or install it. The Software may show a link
to the Whisperstream website's download page at
https://whisperstream.io/download. If you choose that link, the
Software opens it in your web browser. The website, including any
information its download page asks you to provide, is governed by
the website's own terms and privacy notice, and the open-source
edition is governed by its own license terms and privacy policy, not
by this Agreement. Installing that edition is your choice and is
free of charge. Its installer may replace this Software on your
Device, and this Agreement does not promise a way to return to this
version line afterwards. This Agreement continues to govern your use
of this Software and events that occurred before any replacement.
Lanreal does not promise that the website download, the open-source
edition, or its installer will be available at any particular time.

(c) General. By accepting this Agreement, you consent to the
Software performing the update checks described in this Section 10
and applying signed updates that you choose to install. Updates may,
from time to time, modify, add, replace, or remove features of the
Software for product-improvement, business, operational, security,
legal, technical, compatibility, dependency-availability, licensing,
or third-party-service reasons. Except where applicable law requires
otherwise, Lanreal does not promise to release updates or security
patches, follow a maintenance schedule, preserve compatibility with
any operating system, hardware, application, or third-party service,
or keep any feature, integration, or model available. An update may
be necessary for any of the reasons listed above, and an earlier
release may cease to work with third-party services. Declining or
postponing an update does not require Lanreal to maintain or support
the earlier release. Any notice, continued-access option,
termination right, refund, or other remedy that applicable law does
not permit Lanreal to exclude remains available.

11. THIRD-PARTY COMPONENTS
==========================

The Software incorporates Third-Party Components that are licensed
under their own terms. A complete list of all such components,
together with their respective licenses and copyright notices, is
provided in the file THIRD_PARTY_LICENSES.txt bundled with the
Software. It is available in the installed resources directory,
regardless of whether you selected the current-user or all-users
installation mode. The terms of those third-party licenses apply to
those components only and do not modify the terms of this Agreement
as they apply to the Software as a whole, except that Section 27
sets out use restrictions that the license for the Supertonic
Restricted Voice Model requires Lanreal to pass on to you as an
enforceable term of this Agreement.

The following sub-clauses apply to specific Third-Party Components:

(a) LGPL source offer (pygame, pygame-ce, pynput). The Software
incorporates pygame or pygame-ce (each licensed under the GNU Lesser
General Public License, version 2.1 or later) and pynput (licensed
under the GNU Lesser General Public License, version 3). On written
request to support@whisperstream.io, we will provide the complete
corresponding object code of the Software in a form suitable for
relinking (specifically, the Nuitka-generated intermediate object
files and the build instructions necessary to produce a linked
executable), together with any modifications Lanreal has made to
these libraries, to allow you to relink the Software with a modified
version of pygame, pygame-ce, or pynput. This offer is valid for
three (3) years from the date you first download or install the
Software.

(b) Mozilla Public License 2.0 unmodified declaration. Whisperstream
incorporates components licensed under the Mozilla Public License
2.0; these components are used unmodified, and their source is
available at the URLs documented in THIRD_PARTY_LICENSES.txt. The
components in question are: certifi, tqdm, lightningcss (the npm
package and the lightningcss-win32-x64-msvc native binary), and the
Rust crates cssparser, cssparser-macros, dtoa-short, option-ext, and
selectors.

(c) NVIDIA Parakeet TDT 0.6B v3 (CC BY 4.0). Whisperstream uses the
Parakeet TDT 0.6B v3 automatic speech recognition model, developed
by NVIDIA, distributed via Hugging Face under the Creative Commons
Attribution 4.0 International (CC BY 4.0) license. The model is used
unmodified.

(d) r-efi MIT election. The Rust crate r-efi is offered by its
upstream maintainers under a choice of three licenses (Apache-2.0 OR
LGPL-2.1-or-later OR MIT). Lanreal hereby elects to use and
distribute r-efi under the terms of the MIT License. The full text
of the MIT License as it applies to r-efi is reproduced in
THIRD_PARTY_LICENSES.txt.

(e) Supertonic 3 Restricted Voice Model (OpenRAIL-M). The optional
Supertonic model described in Section 27 is distributed by Supertone
Inc. under the BigScience Open RAIL-M License. That license permits
commercial use and redistribution and requires its Attachment A
behavioural use restrictions to be passed on to you as an
enforceable term. Those restrictions are set out in Section 27 and
apply to you only if you download or enable Supertonic. The full
verbatim license text, including Attachment A, is reproduced in
THIRD_PARTY_LICENSES.txt.

(f) Kokoro 82M Voice Model (Apache-2.0). The optional Kokoro Voice
Model described in Section 27 includes the Kokoro 82M FP32 ONNX
model, nine preset voice embeddings, and supporting English frontend
data. The Kokoro model, Misaki lexicons, and PeterReid neural G2P
model are distributed under the Apache License 2.0. Exact MIT
notices and conservative training-source attribution are also
reproduced in THIRD_PARTY_LICENSES.txt. Kokoro is not governed by
OpenRAIL-M, and the use restrictions in Section 27 do not apply to
your use of Kokoro.

(g) Cohere Transcribe 03-2026 (Apache-2.0). If you manually select
Cohere Transcribe, the Software downloads an INT8 ONNX conversion of
the Cohere Transcribe 03-2026 speech-recognition model for on-device
use. Cohere Labs publishes the upstream model under the Apache
License 2.0. The exact upstream revision, conversion and
quantization attribution, modification notice, artifact hashes, and
full Apache License 2.0 text are reproduced in
THIRD_PARTY_LICENSES.txt. The model requires an explicit supported
language, runs entirely on your Device, and is never selected by
Automatic model routing. "Cohere Transcribe" identifies the model
and does not imply that Cohere sponsors, endorses, or is affiliated
with Lanreal or Whisperstream.

12. THIRD-PARTY SERVICES (BRING YOUR OWN KEY)
=============================================

Optional cloud post-processing feature. The Software includes an
optional AI cleanup feature that, when you explicitly enable it, can
route transcribed text through a third-party Large Language Model
service of your choice for post-processing (for example, grammar
cleanup, punctuation normalization, style adjustment, or custom
formatting). This feature is disabled by default. It can also run on
a Large Language Model hosted on your own Device, in which case your
transcribed text does not leave your Device; the terms in this
Section 12 apply to the optional cloud path. The third-party
services the Software supports for the cloud path are enumerated in
Section 6 of the Whisperstream Privacy Policy. When Lockdown Mode
(Section 14) is on, this optional cloud path is disabled and cannot
be used, and AI cleanup runs only on the on-device model. If you
choose to enable the cloud path, the following terms apply.

(a) Your relationship is with the third-party provider. You
acknowledge that your use of any third-party service that you
connect to through the Software is governed by your own agreement
with that provider, including the provider's terms of service,
acceptable use policy, content policy, and privacy policy. Lanreal
is not a party to that agreement, and Lanreal does not assume any
obligation owed to you by the provider. Lanreal is not a reseller,
sub-processor, or agent of any listed provider.

(b) Your credentials, your account, your charges. You represent and
warrant that you have a valid right to use any API key, credential,
or access token you supply to the Software for use with a
third-party provider, and that the use you make of the Software
together with that credential is permitted under your agreement with
the provider. You are responsible for charges imposed by a provider
for requests you authorize through the Software, and for the
provider's rate limits, suspensions, and content-policy enforcement.
If you reasonably believe a material provider charge was caused
solely by a reproducible Software defect, stop using the affected
integration and notify Lanreal promptly with non-content diagnostic
information. Lanreal does not control provider pricing, quotas,
credits, or refunds. Nothing in this paragraph limits a remedy that
cannot lawfully be excluded, and any other recovery is subject to
Section 15.

(c) No warranty of third-party output. Lanreal makes no warranty,
express, implied, or statutory, regarding the availability,
accuracy, latency, reliability, content-policy compliance,
appropriateness, or quality of any output produced by a third-party
service and returned through the Software. Any text that has been
post-processed by a third-party service is, in addition to the
disclaimers in Section 14, explicitly not warranted to be free of
errors, omissions, fabrications, bias, or unintended changes. You
must independently review any such output before relying on it,
including for safety-critical, legal, medical, or financial
purposes. The disclaimers in this Section 12(c) are in addition to,
and do not limit, the general Warranty Disclaimer in Section 14.

(d) No responsibility for the provider. Lanreal does not control and
is not responsible for the availability, security, business
practices, data handling, training-data usage, retention, or any act
or omission of any third-party provider the Software connects to.
Any dispute you may have with a third-party provider (including
disputes about billing, account suspension, content moderation, data
retention, or alleged misuse of submitted content) must be resolved
between you and that provider. Lanreal has no ability to intervene
in, adjudicate, or remediate any such dispute on your behalf.

(e) Compliance with provider terms and applicable law. You agree
that, when using the optional cloud post-processing feature, you
will comply with the selected provider's terms of service and
acceptable use policy, and with all laws and regulations applicable
to the content you submit for post-processing. You must not use the
feature to submit content that you do not have the right to transmit
to the provider, or content whose transmission would violate
applicable law.

(f) Your credentials are not transmitted to Lanreal. Any credential
you supply to the Software for use with a third-party provider is
stored locally on your device as described in Section 6.4 of the
Whisperstream Privacy Policy, and is transmitted only to the
provider you selected, only over HTTPS. Lanreal does not receive,
store, log, or otherwise have access to any credential you supply to
the Software for the optional post-processing feature, and Lanreal
cannot recover a credential on your behalf if you lose it.

(g) Opt-in, opt-out. The optional cloud post-processing feature is
disabled by default in every new installation. The first time you
attempt to enable cloud post-processing with any provider, the
Software will display a blocking consent dialog that identifies the
provider you selected, describes what data will leave your device,
links to the Whisperstream Privacy Policy and to the provider's own
privacy policy, and requires an affirmative click to continue. The
Software will not begin transmitting transcribed text to any
provider until you have completed all required configuration steps
and confirmed the blocking consent dialog. You may enable the
feature, disable it, or delete a stored credential at any time
through the Software's settings interface, as further described in
Section 6.5 of the Whisperstream Privacy Policy. Your use of the
Software without enabling this feature is not subject to this
Section 12.

13. UPSTREAM ATTRIBUTION (WILL HAMPSON / CHIRP, MIT)
====================================================

Whisperstream is built upon portions of source code originally
released under the MIT License by Will Hampson; the MIT License
remains in effect for those portions and its full text is included
in the file LICENSE bundled with the Software. The Whisperstream End
User License Agreement governs Lanreal Technologies Inc.'s
contributions and the bundled product as a whole.

The upstream copyright notice that is preserved in the bundled
LICENSE file reads, verbatim: "Copyright (c) 2025 whamp". You must
not remove or alter that notice, and any redistribution permitted
under the MIT License must continue to reproduce it in accordance
with the terms of that license.

14. WARRANTY DISCLAIMER
=======================

THE SOFTWARE IS PROVIDED "AS-IS" AND "AS-AVAILABLE," WITHOUT
WARRANTY OF ANY KIND, EXPRESS, IMPLIED, OR STATUTORY, INCLUDING BUT
NOT LIMITED TO THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS
FOR A PARTICULAR PURPOSE, ACCURACY, TITLE, NON-INFRINGEMENT, AND
UNINTERRUPTED OR ERROR-FREE OPERATION. LANREAL DOES NOT WARRANT THAT
THE SOFTWARE WILL MEET YOUR REQUIREMENTS, THAT THE OPERATION OF THE
SOFTWARE WILL BE UNINTERRUPTED OR ERROR-FREE, OR THAT DEFECTS IN THE
SOFTWARE WILL BE CORRECTED. THE ENTIRE RISK ARISING OUT OF THE USE
OR PERFORMANCE OF THE SOFTWARE REMAINS WITH YOU.

ASR accuracy disclaimer. Whisperstream uses an automated speech
recognition model that, like all such systems, is inherently
imperfect and may misrecognize words, names, technical terms, or
speech in non-standard accents or noisy environments. The Software
is not warranted to produce accurate transcriptions and must not be
relied upon for safety-critical, legal, medical, or financial
dictation without independent human review. Lanreal is not a
"Covered Entity" or "Business Associate" under the United States
Health Insurance Portability and Accountability Act (HIPAA), and the
Software is not offered as a HIPAA-compliant service. Persons
subject to HIPAA must not use the Software to process Protected
Health Information without independently ensuring their own
compliance with HIPAA's Security and Privacy Rules.

Lockdown Mode disclaimer. The Software offers an optional "Lockdown
Mode" that hardens the Software for privacy-sensitive and
HIPAA-conscious workflows (for example, by running AI cleanup only
on your Device, disabling cloud providers, and locking your
transcript history behind a password), as described in Section 3.6
of the Whisperstream Privacy Policy. Lockdown Mode is designed to
support HIPAA-conscious workflows. It is not a HIPAA certification,
it does not make the Software HIPAA compliant, and it does not
replace your own compliance obligations; the United States
Department of Health and Human Services does not certify or approve
any product as HIPAA compliant. As stated above, Lanreal is not a
Covered Entity or Business Associate. Because Whisperstream's audio,
transcription, history, and AI-cleanup pipeline stays on your Device
in Lockdown Mode and Lanreal never receives that content, no third
party processes it on Lanreal's behalf for that pipeline, and there
is no Business Associate Agreement between you and Lanreal. Turning
Lockdown Mode off and selecting a cloud post-processing provider
under Section 12 would transmit transcribed text to that provider
and change this. Lockdown Mode disables Agent Speak in the Software,
but text you deliver to a destination application and the separate
operation of Claude Code, Codex, or another coding agent remain
outside Lockdown Mode's boundary. You remain responsible for the
administrative, physical, and organizational safeguards that apply
to your own use of the Software.

Lockdown Mode retains bounded, content-safe operational diagnostics
on your Device so you can troubleshoot failures and prepare a useful
bug report. Those diagnostics exclude transcript and voice/audio
content, passwords, recovery codes, keys and license secrets, and
full email addresses. The Software does not upload diagnostics
automatically. A support archive leaves your Device only when you
explicitly save it and then choose to send it or place it in a
location managed by another application. The separate Lockdown
access log is not included in that archive. See Section 3.4 of the
Whisperstream Privacy Policy.

Lockdown Mode data loss. If you enable the Lockdown Mode transcript
password, your password and your one-time recovery code are held
only on your Device and are never transmitted to Lanreal. If you
lose both your password and your recovery code, your transcript
history cannot be unlocked through the Software, and Lanreal cannot
unlock or recover it for you. There is no Lanreal-side backup. The
only available reset removes the locked transcript history through
ordinary database and file deletion and starts over; it cannot
restore the old records or securely overwrite the storage device. To
the maximum extent permitted by applicable law, Lanreal is not
liable for any loss of, or inability to access, transcript history
resulting from a lost or forgotten password or recovery code.

Text-injection compatibility disclaimer. Whisperstream delivers text
to the focused or pinned application using clipboard paste or
OS-level keystroke simulation. Some applications, security software,
elevated/administrator windows, and User Account Control prompts may
refuse to receive simulated keystrokes. The Software is not
warranted to function correctly in such contexts.

15. LIMITATION OF LIABILITY
===========================

TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT SHALL
LANREAL, ITS DIRECTORS, OFFICERS, EMPLOYEES, AGENTS, CONTRACTORS, OR
LICENSORS BE LIABLE TO YOU OR TO ANY THIRD PARTY FOR ANY INDIRECT,
INCIDENTAL, CONSEQUENTIAL, SPECIAL, EXEMPLARY, PUNITIVE, OR SIMILAR
DAMAGES, INCLUDING, WITHOUT LIMITATION, LOST PROFITS, LOST REVENUE,
LOST DATA, BUSINESS INTERRUPTION, LOSS OF GOODWILL, OR THE COST OF
SUBSTITUTE GOODS OR SERVICES, ARISING OUT OF OR IN CONNECTION WITH
THIS AGREEMENT OR YOUR USE OF OR INABILITY TO USE THE SOFTWARE,
WHETHER BASED ON WARRANTY, CONTRACT, TORT (INCLUDING NEGLIGENCE),
STRICT LIABILITY, OR ANY OTHER LEGAL THEORY, AND WHETHER OR NOT
LANREAL HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

LANREAL'S TOTAL CUMULATIVE LIABILITY TO YOU UNDER OR IN CONNECTION
WITH THIS AGREEMENT, FROM ALL CAUSES OF ACTION AND UNDER ALL
THEORIES OF LIABILITY, SHALL NOT EXCEED THE LICENSE FEE PAID BY YOU
AT CHECKOUT FOR THE PRO LICENSE GIVING RISE TO THE CLAIM, OR, WHERE
NO SUCH FEE WAS PAID, ZERO.

Notwithstanding the foregoing, nothing in this Section 15 shall
limit (a) either party's liability for fraud or fraudulent
misrepresentation, (b) Lanreal's liability for death or personal
injury caused by Lanreal's negligence, or (c) any liability that
cannot be excluded or limited by applicable law.

The limitations in this Section 15 apply to the fullest extent
permitted by applicable law, including to claims arising from the
optional cloud post-processing feature described in Section 12, and
survive termination of this Agreement. Sections 6 (Prohibited Use),
7 (AI/ML Training), 11(a) (LGPL source offer), 12 (Third-Party
Services, with respect to your obligations and Lanreal's
disclaimers), 13 (Upstream Attribution), 14 (Warranty Disclaimer),
15 (Limitation of Liability), 16 (Governing Law and Jurisdiction),
17 (Consumer Protection Non-Derogation), 18 (Language), 19 (Entire
Agreement and Related Matters), 20 (Amendments), 21 (Force Majeure),
22 (Notices), 23 (End of Life), 24 (Export Compliance and
Sanctions), 25 (Age and Capacity), and 27 (Optional Voice Feature
and Supertonic Use Restrictions) also survive termination of this
Agreement.

16. GOVERNING LAW AND JURISDICTION
==================================

This Agreement is governed by and construed in accordance with the
laws of the Province of Ontario and the laws of Canada applicable
therein, without regard to conflict-of-laws principles. The United
Nations Convention on Contracts for the International Sale of Goods
does not apply to this Agreement.

The parties consent to the exclusive jurisdiction of the courts of
the Province of Ontario for any dispute arising out of or relating
to this Agreement.

Nothing in this Section 16 prevents you, as a consumer, from
bringing proceedings in the courts of any jurisdiction where
mandatory consumer protection law gives you an indefeasible right to
bring proceedings in those courts.

17. CONSUMER PROTECTION NON-DEROGATION
======================================

Nothing in this Agreement excludes or limits any non-waivable
statutory rights you may have under the consumer protection laws of
your jurisdiction. Where mandatory consumer protection law provides
you with greater rights than this Agreement, those mandatory rights
prevail to the extent of the inconsistency.

18. LANGUAGE
============

This Agreement has been drafted in English at the express request of
the parties. Les parties ont expressement exige que la presente
convention soit redigee en langue anglaise.

If you are a consumer resident in the Province of Quebec, Lanreal
will provide you, promptly and without charge, with a French
translation of this Agreement on request by emailing
support@whisperstream.io. Lanreal aims to respond to any such
request within five (5) business days. Without limiting any rights
you may have under Section 17, if a court or the Office quebecois de
la langue francaise determines that any provision of this Agreement
is inopposable against you because a French translation was not made
available to you before you were bound by this Agreement, that
provision will be treated as inopposable against you and the
remainder of this Agreement will continue to apply to the fullest
extent permitted by law. This paragraph, together with Section 17
(Consumer Protection Non-Derogation), is intended to respect the
language obligations of Quebec consumer law, including the Charter
of the French Language as amended.

19. ENTIRE AGREEMENT AND RELATED MATTERS
========================================

Entire Agreement. This Agreement, together with the documents
expressly incorporated by reference (including
THIRD_PARTY_LICENSES.txt, the bundled LICENSE file, and the
Whisperstream Privacy Policy), constitutes the entire agreement
between you and Lanreal with respect to the Software, and supersedes
all prior or contemporaneous oral or written communications,
proposals, and representations with respect to the Software.

Customer documents. A purchase order, procurement portal, vendor
onboarding form, security questionnaire, or similar customer
document may record administrative and commercial details such as
the purchaser, license quantity, price, and billing contact. Lanreal
rejects any additional or conflicting legal terms in such a
document. Processing an order or payment, issuing or activating a
License Key, fulfilling a purchase, completing or responding to a
customer form, providing a checkout link, or remaining silent does
not accept those terms. Only a separate written agreement that
expressly identifies this Agreement and each provision it changes
and is signed by an authorized officer of Lanreal may modify this
Agreement.

Support and informational materials. Except for obligations
expressly stated in this Agreement or required by applicable law,
any support Lanreal provides is discretionary and subject to
availability. Unless a separate written agreement signed by an
authorized officer of Lanreal expressly provides otherwise, Lanreal
does not commit to uptime, availability, maintenance, response or
resolution times, compatibility, deployment assistance,
configuration, troubleshooting, or professional services.
Documentation, security materials, questionnaire responses, and
other product materials describe the Software as of the date stated
or, if no date is stated, the date Lanreal provided them. They are
informational descriptions, not warranties, service levels,
certifications, support obligations, or promises to maintain a
feature or practice. Nothing in this paragraph limits responsibility
for fraud, fraudulent misrepresentation, or a right that applicable
law does not permit Lanreal to exclude.

Severability. If any provision of this Agreement is held by a court
of competent jurisdiction to be invalid, illegal, or unenforceable,
that provision shall be enforced to the maximum extent permissible
so as to effect the intent of the parties, and the remaining
provisions of this Agreement shall remain in full force and effect.

No Waiver. No failure or delay by Lanreal in exercising any right,
power, or privilege under this Agreement shall operate as a waiver
thereof, nor shall any single or partial exercise of any such right,
power, or privilege preclude any other or further exercise thereof,
or the exercise of any other right, power, or privilege.

Assignment. You may not assign or transfer this Agreement or any of
your rights or obligations under it, by operation of law or
otherwise, without Lanreal's prior written consent. Lanreal will not
unreasonably withhold that consent for personal assignments between
family members or transfers arising from death or incapacity. In
other cases, Lanreal may grant or withhold consent in its
discretion, subject to rights that applicable law does not permit
Lanreal to exclude. Any attempted assignment in violation of this
provision is void. Lanreal may assign this Agreement without your
consent to an affiliate or in connection with a merger, acquisition,
reorganization, or sale or transfer of the Software, a product line,
a business, or related assets, provided that the assignee assumes
Lanreal's obligations under this Agreement and the assignment does
not reduce any consumer protection that applicable law does not
permit Lanreal to exclude. Lanreal will provide notice of an
assignment where applicable law requires it.

Data on uninstall. Uninstalling the Software does not automatically
delete your local configuration, logs, usage statistics, saved
transcript history (your encrypted transcribed text and any saved
audio), stored credentials, legal-acceptance records, or license and
trial records left by earlier versions from your Windows user
profile and current-user registry. You may delete
%LOCALAPPDATA%\Whisperstream\ and the Whisperstream value under
HKCU\Software\Whisperstream manually after uninstalling, or delete
transcript history from within the Software first. If you selected a
custom model storage location, its models and local-llm directories
also remain and must be deleted separately. For information about
personal data held by Polar Software Inc. about your purchase or
refund, see the Whisperstream Privacy Policy and Polar's own privacy
policy.

20. AMENDMENTS
==============

Lanreal may publish an updated Agreement from time to time. When we
do, we will change the version number and effective date at the top
of this document. The version number generally tracks the version of
the Software with which the Agreement is bundled and is an
identifier only. The current bundled version and the most recent
published version at https://whisperstream.io/eula/private may
differ if you have not installed the latest release.

Changes requiring affirmative acceptance. A change that affects your
rights or obligations does not bind you unless Lanreal provides
clear notice, a plain-language summary, and access to the complete
updated Agreement, and you then affirmatively accept it through the
Software or another clear click-through process. Posting updated
terms, silence, closing a notice, or continued use does not
constitute acceptance.

Lanreal may require acceptance before you use an updated release or
a new optional feature. If you decline, you must not use that
updated release or feature. The Agreement you previously accepted
continues to govern your lawful use of any earlier release that
remains available to you. Once accepted, the updated Agreement
governs your use of that release and events occurring after
acceptance, including future updates, support, and discontinuation.
No amendment applies retroactively or releases any claim, refund
entitlement, or remedy that accrued from events occurring before it
became binding.

Administrative corrections. Lanreal may correct typographical
errors, broken links, formatting, or other administrative details
that do not change the meaning of this Agreement. Those corrections
may take effect when published without a new acceptance.

Mandatory rights. Lanreal will provide any notice, language,
refusal, cancellation, refund, continued-access, or other protection
required by applicable law. Nothing in this Section limits the
non-waivable consumer rights preserved by Section 17.

Acceptance record. When you accept an updated Agreement, the
Software appends exact acceptance evidence to
%LOCALAPPDATA%\Whisperstream\acceptance.json, including the
Agreement version and hash, acceptance time, installed Software
version, and acceptance event type. The file also exposes the latest
acceptance as its current record, but a later acceptance preserves
the earlier evidence in the same local ledger. Declining, closing,
or deferring the prompt does not record acceptance. The Software may
ask again the next time you launch it. If you decline an updated
Agreement, the Software may record that decision locally and suspend
automatic update checks for up to thirty (30) days; you can still
check for updates manually.

21. FORCE MAJEURE
=================

To the maximum extent permitted by law, Lanreal is not responsible
for a failure or delay to the extent caused by circumstances beyond
its reasonable control, including natural disasters, acts of
government, internet or telecommunications failures, power outages,
labour disputes, malicious attacks, or failures of third-party
providers. No refund or termination right arises solely from such a
circumstance except as expressly provided in Section 8 or required
by applicable law.

22. NOTICES
===========

Any formal notice required or permitted under this Agreement must be
in writing and delivered by email to the addresses below. Notices to
Lanreal must be sent to support@whisperstream.io. Notices to you
will be sent to the email address associated with your license
purchase through Polar, if available, or posted prominently on the
Whisperstream website at https://whisperstream.io. A notice
describing an individual breach under Section 3 will be sent to the
last email address you provided or delivered through an
in-application notice tied to the affected installation.

23. END OF LIFE
===============

For product-improvement, business, operational, security, legal,
technical, compatibility, dependency-availability, licensing, or
third-party-service reasons, Lanreal may discontinue development,
support, updates, downloads, activation services, distribution, or
any feature of the Software. Lanreal may, at its discretion, provide
advance notice or transition measures, but, except where applicable
law requires otherwise, does not promise any particular notice
period, unlocked build, replacement service, source code, continued
download or activation, or refund solely because of a change or
discontinuation. This version of the Software does not depend on
activation services.

The license granted in Section 3 has no fixed contractual
expiration. That phrase describes the duration of that license right
and does not guarantee that the Software, any feature, downloads,
activation, updates, support, third-party services, or compatibility
will remain available or continue to operate indefinitely. This
Section does not limit the refund right in Section 8 or the source
code rights or offers in Section 11. Any notice, continued-access
option, termination right, refund, or other remedy that applicable
law does not permit Lanreal to exclude remains available.

Major version upgrades. Lanreal may offer a major new version as a
separate product, which may require a new purchase. The license in
Section 3 covers this release and any update in the same major
version line that Lanreal elects to make available. It does not
entitle you to an update or require Lanreal to keep an update or
download available. Major new versions are not included unless
Lanreal expressly states otherwise. The open-source edition
described in Section 10(b) is a separate product under its own
terms. It is not an update under this Agreement, and linking to it
does not make this Agreement apply to it.

24. EXPORT COMPLIANCE AND SANCTIONS
===================================

You may not export, re-export, or import the Software or any direct
product of the Software in violation of any applicable law,
including the United States Export Administration Regulations, the
Canadian Special Economic Measures Act, or the laws of any country
to which the Software is exported. You represent and warrant that
you (a) are not located in, under the control of, or a national or
resident of any country to which the export of the Software is
prohibited by applicable law, and (b) are not listed on any United
States Government list of prohibited or restricted parties
(including the Specially Designated Nationals list maintained by
OFAC, the Entity List, or the Denied Persons List), nor on any
equivalent Canadian, EU, or UK sanctions list. You agree to comply
with all applicable export and sanctions laws in your use of the
Software.

25. AGE AND CAPACITY
====================

You represent and warrant that you are at least the age of majority
in the jurisdiction in which you reside (and in any event at least
18 years of age), or that you are using the Software with the
consent and supervision of a parent or legal guardian who has
reviewed and accepted this Agreement on your behalf. Lanreal does
not knowingly direct the Software to children under thirteen (13)
years of age and does not knowingly collect personal information
from children under thirteen. If you are a parent or guardian and
you become aware that a child under thirteen is using the Software
in a manner inconsistent with the terms of this Agreement, contact
support@whisperstream.io and Lanreal may provide available
instructions for uninstalling the Software and removing locally
stored data.

26. CONTACT
===========

For any notices, refund requests, LGPL source-code requests, or
other communications relating to this Agreement, please contact
Lanreal at:

Lanreal Technologies Inc. 18 King Street East, Suite 1400 Toronto,
ON M5C 1C4 Canada Email: support@whisperstream.io

27. OPTIONAL VOICE FEATURE AND SUPERTONIC USE RESTRICTIONS
==========================================================

Optional voice feature. The Software includes an optional voice
feature that reads an artificial-intelligence coding assistant's
text replies aloud on your Device using a selected preset synthetic
voice. This feature is disabled by default. The speech models
available to the feature (the "Voice Models") are not bundled with
the installer and are not part of the "Software" as defined in
Section 2. If, and only if, you enable the feature, the Software
downloads the Voice Model or Voice Models needed for your selected
or automatically resolved engine to your Device at your direction,
as described in Section 9(e). The Voice Models are not voice-cloning
or speaker-adaptation tools. The Software does not accept a user's
voice samples and provides no capability to train, clone, mix,
import, or adapt a Voice Model.

When this Section applies. This Section 27 applies to you only if
you choose to download or enable the optional voice feature. The
OpenRAIL-M use restrictions below apply only if you download,
enable, or use Supertonic 3 (the "Restricted Voice Model"). They do
not apply to Kokoro 82M (the "Kokoro Voice Model"). If you do not
download or enable the Restricted Voice Model, the use restrictions
in this Section 27 impose no obligation on you. Nothing in this
Section 27 narrows the license granted to you in Section 3, limits
the permitted uses described in Section 5, or adds to the
prohibitions in Section 6 in respect of the Software itself or the
Kokoro Voice Model.

Kokoro attribution and governing license. The Kokoro Voice Model is
Kokoro 82M, distributed under the Apache License 2.0. Its pinned
source is published at https://huggingface.co/hexgrad/Kokoro-82M.
The optional archive also includes Apache-2.0 Misaki lexicons and a
PeterReid neural G2P model, exact MIT notices for NLTK tagger
material and ONNX conversion lineage, and conservative
training-source attribution. The governing texts and notices are
delivered in THIRD_PARTY_LICENSES.txt. OpenRAIL-M does not govern
the Kokoro Voice Model.

Supertonic attribution and governing license. The Restricted Voice
Model is Supertonic 3, Copyright (c) 2026 Supertone Inc.,
distributed by Supertone Inc. under the BigScience Open RAIL-M
License ("OpenRAIL-M"). Its model card is published at
https://huggingface.co/Supertone/supertonic-3. OpenRAIL-M permits
commercial use and redistribution, and requires in return that its
Attachment A use restrictions be passed on to you as an enforceable
term. That is the reason this Section 27 exists. The full verbatim
text of OpenRAIL-M, including Attachment A, is delivered to you in
the file THIRD_PARTY_LICENSES.txt bundled with the Software and
described in Section 11. Lanreal does not modify the Restricted
Voice Model and distributes it as received.

Your Supertonic use restrictions. If you download, enable, or use
the Restricted Voice Model, you agree that you will not use the
Restricted Voice Model, and will not permit any third party to use
the Restricted Voice Model:

(a) in any way that violates any applicable national, federal,
state, local, or international law or regulation;

(b) for the purpose of exploiting, harming, or attempting to exploit
or harm minors in any way;

(c) to generate or disseminate verifiably false information or
content with the purpose of harming others;

(d) to generate or disseminate personally identifiable information
that can be used to harm an individual;

(e) to generate or disseminate information or content (for example,
images, code, posts, or articles), or to place that information or
content in any context (for example, a bot generating social media
posts), without expressly and intelligibly disclaiming that the
information or content is machine generated;

(f) to defame, disparage, or otherwise harass others;

(g) to impersonate or attempt to impersonate any person or entity,
including by means of synthetic media commonly described as
deepfakes;

(h) for fully automated decision-making that adversely impacts an
individual's legal rights or otherwise creates or modifies a
binding, enforceable obligation;

(i) for any use intended to, or having the effect of, discriminating
against or harming individuals or groups based on online or offline
social behaviour or known or predicted personal or personality
characteristics;

(j) to exploit any of the vulnerabilities of a specific group of
persons based on their age or their social, physical, or mental
characteristics, in order to materially distort the behaviour of a
person belonging to that group in a manner that causes or is likely
to cause that person or another person physical or psychological
harm;

(k) for any use intended to, or having the effect of, discriminating
against individuals or groups based on legally protected
characteristics or categories;

(l) to provide medical advice or to interpret medical results; or

(m) to generate or disseminate information for use in the
administration of justice, law enforcement, immigration, or asylum
processes, including predicting that an individual will commit fraud
or a crime, whether by text profiling, by drawing causal
relationships between assertions made in documents, or by
indiscriminate or arbitrarily targeted use.

Incorporation of Attachment A. The restrictions in paragraphs (a)
through (m) above are Lanreal's rendering of the Attachment A use
restrictions of OpenRAIL-M, included in this Agreement as an
enforceable provision as OpenRAIL-M requires. The verbatim
Attachment A text delivered in THIRD_PARTY_LICENSES.txt is
incorporated into this Section 27 by reference. If any restriction
in that verbatim Attachment A is broader than the corresponding
paragraph above, the verbatim text governs your use of the
Restricted Voice Model to the extent of the difference. Nothing in
this Section 27 narrows any Attachment A restriction.

Termination of the Restricted Voice Model right; survival. Your
right to use the Restricted Voice Model is conditional on your
compliance with this Section 27. If you breach this Section 27, your
right to use the Restricted Voice Model terminates automatically and
without notice, and you must stop using the Restricted Voice Model
and delete it from your Device. Termination of your right to use the
Restricted Voice Model under this Section does not by itself
terminate the license granted to you in Section 3, which is governed
by Section 3, and does not terminate your right to use the Kokoro
Voice Model under its Apache-2.0 license. This Section 27 survives
termination or expiry of this Agreement, and survives your removal
of the Restricted Voice Model or of the Software from your Device.

No third-party beneficiary obligations; disclaimers apply. Supertone
Inc. and the other creators and licensors of the Voice Models are
not parties to this Agreement and owe you no obligation under it.
The warranty disclaimer in Section 14 and the limitation of
liability in Section 15 apply to the Voice Models and to any audio
they produce. Questions about this Section may be sent to
support@whisperstream.io.