Credits & Acknowledgements.

Doza Assist stands on remarkable open-source software, AI models, and storytelling research. This page mirrors the NOTICES document that ships inside every copy of the app.

Doza Assist is a proprietary commercial product. It builds on, and depends on, third-party open-source software and AI models. This document collects the required attributions and license notices for those components.

The proprietary portions of Doza Assist are governed by the proprietary license that accompanies the app. The third-party components below remain governed by their respective licenses, which take precedence over Doza Assist's proprietary license for those specific files and works.

Table of Contents

1. Doza Assist (open-source core)
2. AI Models — Gemma 4, OpenAI Whisper, Parakeet TDT
3. Ollama runtime
4. Electron framework and runtime tooling
5. Python runtime dependencies
6. JavaScript runtime dependencies
7. Build-time tools (electron-builder, signing, FFmpeg build chain)
8. FFmpeg (LGPL — bundled binary)
9. Storytelling foundation

01Doza Assist (open-source core)

Repository: https://github.com/DozaVisuals/doza-assist Copyright: © Doza Visuals LLC and contributors License: MIT License

The open-source Doza Assist project provides the Python/Flask backend. It ships inside the app as plain .py source files under Contents/Resources/app/, copied at build time via electron-builder's extraResources. The MIT license text:

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.

02AI Models

2.1 Gemma 4

• Copyright: Google LLC
• License: Apache License, Version 2.0 — https://www.apache.org/licenses/LICENSE-2.0
• Model card: https://huggingface.co/collections/google/gemma-4-release
• Terms: https://ai.google.dev/gemma/terms

Gemma 4 weights are not bundled in this product. They are downloaded at runtime by the user via Ollama (ollama pull gemma4:<variant>). The variant is auto-selected based on the user's hardware (see model_config.py in the OSS core).

Use of Gemma 4 is subject to the Apache 2.0 license and any additional usage policies published by Google.

2.2 OpenAI Whisper

• Copyright: OpenAI
• License: MIT License — https://github.com/openai/whisper/blob/main/LICENSE

Whisper is the non-English transcription fallback engine. The openai-whisper Python package is bundled in the app's python-site-packages/ directory. Whisper model weights are downloaded at runtime on first use; the weights are not bundled.

2.3 Parakeet TDT 0.6B v2 (MLX Community conversion)

• Original model: nvidia/parakeet-tdt-0.6b-v2 — © NVIDIA Corporation
• Converted by: MLX Community contributors — https://huggingface.co/mlx-community
• License: Creative Commons Attribution 4.0 International (CC-BY-4.0) — https://creativecommons.org/licenses/by/4.0/
• Hugging Face: https://huggingface.co/mlx-community/parakeet-tdt-0.6b-v2

This model is the primary English transcription engine on Apple Silicon hosts. Weights are downloaded at runtime via the parakeet-mlx package from Hugging Face. They are not bundled in this product.

Attribution: Original model by NVIDIA; MLX conversion by the MLX Community.

03Ollama

• Copyright: Ollama, Inc.
• License: MIT License — https://github.com/ollama/ollama/blob/main/LICENSE

The Ollama runtime binary (v0.23.2, obtained from the official GitHub releases) is bundled at Contents/Resources/bin/ollama and run as a managed subprocess for local LLM inference. Ollama is MIT-licensed; the license text is available at the URL above. The bundled instance's model storage is redirected into the app's own support directory and never touches a user-installed Ollama.

04Electron framework and runtime tooling

Doza Assist is built on the Electron framework.

• Copyright: OpenJS Foundation and Electron contributors (electron); electron-builder authors; Loopline Systems (osx-sign).
• Electron repository: https://github.com/electron/electron
• License URL: https://opensource.org/license/mit

Electron's binary distribution embeds Chromium and Node.js. Chromium is licensed under a combination of BSD-3-Clause and other permissive licenses; Node.js is MIT. Their full attribution files ship inside the Electron Framework bundle at Doza Assist.app/Contents/Frameworks/Electron Framework.framework/Resources/LICENSES.chromium.html and similar paths, and are not duplicated here.

electron and electron-builder are devDependencies used at build time. The electron runtime is then bundled inside the .app. electron-updater is the only direct runtime npm dependency that ships in production; it lives at app.asar/node_modules/electron-updater/ after packaging.

Transitive npm dependencies of electron-updater (notably js-yaml, lodash.escaperegexp, semver) are licensed under MIT, BSD, and ISC. A full list can be regenerated with npm ls --production and npm-license-checker.

05Python runtime dependencies

These packages are bundled inside the app at Contents/Resources/python-site-packages/, installed at build time by scripts/prep-build.sh against the bundled CPython interpreter. See BUNDLE_MANIFEST.json in Contents/Resources/ for the exact resolved versions of the full bundled stack (torch, pyannote.audio, openai-whisper, parakeet-mlx, …). The Python source itself ships as plain .py files under Doza Assist.app/Contents/Resources/app/, copied via electron-builder's extraResources from the OSS core repo.

5.1 BSD-3-Clause license text (applies to Pallets and lxml above)

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

1. Redistributions of source code must retain the above copyright notice,
   this list of conditions and the following disclaimer.

2. Redistributions in binary form must reproduce the above copyright notice,
   this list of conditions and the following disclaimer in the documentation
   and/or other materials provided with the distribution.

3. Neither the name of the copyright holder nor the names of its contributors
   may be used to endorse or promote products derived from this software
   without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.

Copyright holders: Pallets project (Flask, Werkzeug, Jinja2, MarkupSafe, click, itsdangerous), lxml authors, idna authors.

5.2 MPL-2.0 (certifi)

certifi is licensed under the Mozilla Public License 2.0. Full text: https://www.mozilla.org/en-US/MPL/2.0/

certifi is a CA bundle data file packaged for Python; redistribution is permitted under MPL 2.0 without obligating Doza Assist to MPL terms.

5.3 Apache 2.0 license note

requests and packaging are licensed under the Apache License 2.0. Full text and conditions at https://www.apache.org/licenses/LICENSE-2.0. Their NOTICE files are reproduced here:

• requests has no NOTICE file beyond the LICENSE text.
• packaging has no NOTICE file beyond the LICENSE text.

06JavaScript runtime dependencies

The Electron renderer is intentionally minimal — splash.html, welcome.html, and theme.css poll the local Flask backend on a chosen port. There are no frontend framework dependencies (no React, no Vue, etc.).

The only direct runtime npm dependency that ships in production is electron-updater (see Section 4). Its transitive dependencies are listed in package-lock.json; all are MIT, BSD, or ISC licensed.

electron and electron-builder are devDependencies — used to package and produce the signed .app. The Electron *runtime* itself is bundled into the app; the electron-builder *tool* is not.

07Build-time tools

These tools produce build artifacts but are NOT bundled inside Doza Assist.app in any form that creates licensing obligations on the shipped product.

7.1 electron-builder

• License: MIT
• URL: https://github.com/electron-userland/electron-builder

Used to package the Electron app, copy extraResources, sign Mach-O binaries, and produce the DMG. The electron-builder tool itself is not bundled into the production app; only its output (a signed .app) ships.

7.2 @electron/osx-sign and @electron/notarize

• License: BSD-2-Clause (osx-sign), MIT (notarize)
• URLs: https://github.com/electron/osx-sign, https://github.com/electron/notarize

Pulled in transitively by electron-builder for codesigning and Apple notarization. Not bundled into the production app.

7.3 FFmpeg build chain

The bundled FFmpeg binaries (Section 8) are produced by scripts/build_ffmpeg.sh from upstream FFmpeg sources at build time. The build script itself is not part of the shipped app. Apple Clang and macOS SDK are used as the toolchain; no third-party build dependencies are required.

08FFmpeg

• Project: FFmpeg — https://ffmpeg.org
• Version bundled: 8.1.1 ("Hoare", released 2026-05-04)
• Copyright: © FFmpeg developers
• License: GNU Lesser General Public License, version 2.1 or later (LGPL-2.1-or-later)
• License text: https://www.gnu.org/licenses/old-licenses/lgpl-2.1.html and reproduced verbatim at Contents/Resources/ffmpeg/LICENSE inside the application bundle.

What is bundled

Doza Assist bundles two binaries built from FFmpeg sources, along with their dynamically-linked LGPL libraries:

• ffmpeg — used by the open-source core for audio extraction (transcribe.py) and FCPXML timeline audio composition (doza_assist/fcpxml/timeline_audio.py).
• ffprobe — used by the open-source core for video container metadata reads (frame rate and resolution detection in app.py and exporters/media_probe.py).

License compliance posture

The bundled FFmpeg binaries are built **without any GPL or non-free components**. Specifically:

• Built with --disable-gpl and --disable-nonfree.
• No --enable-libx264, --enable-libx265, --enable-libfdk-aac, or any other GPL/non-free third-party library.
• Audio-only codec set, all FFmpeg-native implementations with no external codec libraries: AAC (incl. LATM), AC-3/E-AC-3, MP3/MP2, FLAC, ALAC, Opus, Vorbis, WMA v1/v2, AMR-NB/WB, DTS (DCA), SMPTE 302M, and the PCM matrix. Encoding is limited to uncompressed PCM (pcm_s16le, pcm_f32le) — no lossy encoder of any kind is present.
• Built with --enable-shared --disable-static. The binaries link dynamically against the LGPL libraries (libavcodec, libavformat, libavfilter, libavutil, libswresample, libavdevice).

This dynamic-linking posture preserves the end user's LGPL right to substitute a modified version of any FFmpeg library without relinking Doza Assist itself.

Source code and build script availability

Per LGPL 2.1 § 6, the FFmpeg 8.1.1 source tarball, the exact configure flags used, and the build script that produces a functionally equivalent binary are published at a permanent URL:

• https://appcast.doza.ai/sources/ffmpeg-8.1.1/ffmpeg-8.1.1.tar.xz — the unmodified upstream source tarball (SHA-256 b6863adde98898f42602017462871b5f6333e65aec803fdd7a6308639c52edf3).
• https://appcast.doza.ai/sources/ffmpeg-8.1.1/build_ffmpeg.sh — the build script, verbatim.
• https://appcast.doza.ai/sources/ffmpeg-8.1.1/SOURCE_INFO.md — the companion rebuild document.

The build script and SOURCE_INFO.md also ship inside the app bundle at Contents/Resources/ffmpeg/, next to the binaries and the LGPL license text, so recipients of the binary hold the rebuild instructions directly.

These artifacts will remain available for at least three years from the date of distribution of any release that bundles these binaries.

Modifications to upstream

None. No patches were applied to the FFmpeg source tree. The build is a clean configure / make / make install of unmodified upstream source. The only "modification" is a strip of debug and local symbols from the resulting binaries and dylibs, which does not alter functionality.

09Storytelling foundation

The storytelling guidance built into Doza Assist (storytelling-master-v1.md) is original instructional writing by Doza Visuals LLC, created to guide the local language model that powers the app's clip-selection and narrative-reordering features.

It draws on long-established, widely taught principles of film editing, screenwriting, and documentary craft as described in published books, lectures, and interviews. Story frameworks and editing methods are ideas and systems of practice, which copyright does not protect; the document does not reproduce protected creative expression. No affiliation with, or endorsement by, any author, practitioner, publication, or broadcaster is implied.

If you believe anything in this product warrants additional attribution or correction, contact [email protected].

*Last updated: 2026-06-10.*