Third-Party Notices
NotesXML
Copyright 2026 IWV Digital Solutions LLC
This product includes software and/or models developed by third parties. Each component retains its original license.
1. llama.cpp
| Repository | github.com/ggerganov/llama.cpp |
| Author | Georgi Gerganov and contributors |
| License | MIT License |
| Used for | Local LLM inference engine (llama-server binary on Electron, JNI library on Android) powering AI features: summarize, translate, polish, AI chat, vision OCR, transcription enhancement |
2. PDF.js
| Repository | github.com/mozilla/pdf.js (Mozilla) |
| License | Apache License 2.0 |
| Used for | PDF viewing, text extraction, and page rendering for PDF Notes and PDF annotation |
3. pdf-lib
| Repository | github.com/Hopding/pdf-lib |
| License | MIT License |
| Used for | PDF creation, annotation serialization (ISO 32000), and Print-to-PDF output |
4. jsPDF
5. JSZip
| Repository | github.com/Stuk/jszip |
| License | MIT License |
| Used for | ZIP archive creation and extraction for ENEX import and backup/export operations |
6. Marked
7. Pako
| Repository | github.com/nodeca/pako |
| License | MIT License |
| Used for | Zlib compression/decompression for NXL file format and PDF stream handling |
8. UTIF.js
9. heic2any
10. whisper.cpp
11. Electron
12. FFmpeg Kit
13. mammoth.js
| Repository | github.com/mwilliamson/mammoth.js |
| Author | Michael Williamson |
| License | BSD 2-Clause License |
| Used for | DOCX (Microsoft Word) document import and text extraction for AI features |
14. SheetJS (xlsx)
| Repository | sheetjs.com |
| Author | SheetJS LLC |
| License | Apache License 2.0 |
| Used for | XLSX / spreadsheet import and text extraction for AI features |
15. Arm KleidiAI
| Repository | github.com/ARM-software/kleidiai |
| Author | Arm Limited |
| License | Apache License 2.0 |
| Used for | ARM i8mm matmul acceleration kernels compiled into the Android llama.cpp build |
16. Piper TTS (with espeak-ng and ONNX Runtime)
| Repository | github.com/rhasspy/piper |
| License | MIT License (Piper itself) |
| Used for | On-device text-to-speech (Read Aloud and AI voice output) on Windows and Linux. Voice models are downloaded on demand and verified against pinned SHA-256 digests. |
The Piper runtime is distributed together with two companion components that NotesXML also redistributes unmodified:
| espeak-ng | github.com/espeak-ng/espeak-ng — GNU General Public License v3.0. Used by Piper solely for text phonemization. NotesXML redistributes the unmodified espeak-ng library; the complete corresponding source code is available from the repository linked here, and we will provide it on request (legal@iwvdigitalsolutions.com). |
| ONNX Runtime | github.com/microsoft/onnxruntime (Microsoft) — MIT License. Neural network inference engine executing the Piper voice models. |
17. Microsoft Visual C++ Runtime (app-local)
| Provider | Microsoft Corporation |
| Components | vcruntime140.dll, vcruntime140_1.dll, msvcp140.dll (from the Microsoft Visual C++ Redistributable, deployed app-locally beside each bundled engine) |
| License | Redistributed under the Microsoft Visual Studio license terms for REDIST-listed files |
| Used for | C/C++ runtime support for the bundled Windows inference engines (llama-server, whisper, Piper), so AI, dictation and read-aloud features work on every Windows PC without requiring a separate Visual C++ Redistributable installation (added in Beta 5.142.58) |
AI Models
NotesXML downloads AI models on demand from Hugging Face. Each model has its own license:
| Model | Provider | License |
| LFM2.5 VL 1.6B (vision-capable) — free tier, recommended default | Liquid AI | LFM Open License v1.0 — see the note below |
| Ministral 3 3B Instruct (vision-capable) — free tier | Mistral AI | Apache License 2.0 |
| Gemma 4 E2B (vision-capable) | Google DeepMind | Apache License 2.0 |
| LFM2.5 VL 3B (vision-capable) | Liquid AI | LFM Open License v1.0 |
| Gemma 4 E4B (vision-capable) | Google DeepMind | Apache License 2.0 |
| Ministral 3 8B Instruct (vision-capable) | Mistral AI | Apache License 2.0 |
| Mellum2 12B-A2.5B Instruct (MoE; coding) | JetBrains s.r.o. | Apache License 2.0 |
| Gemma 4 12B (vision-capable) | Google DeepMind | Apache License 2.0 |
| Ministral 3 14B Instruct (vision-capable) | Mistral AI | Apache License 2.0 |
| GPT-OSS 20B (MoE) — desktop only | OpenAI | Apache License 2.0 |
| Devstral Small 2 24B (coding; vision-capable) | Mistral AI | Apache License 2.0 |
| Gemma 4 26B-A4B MoE (vision-capable) — desktop only | Google DeepMind | Apache License 2.0 |
| Muse Glimmer 30B (vision-capable) — desktop only | Meta Superintelligence Labs | Apache License 2.0 |
| Nemotron 3.5 Lightning 30B-A3B (MoE) — desktop only | NVIDIA Corporation | OpenMDW License Agreement v1.1 |
| Gemma 4 31B (Dense) (vision-capable) — desktop only | Google DeepMind | Apache License 2.0 |
| Laguna XS 2.1 33B-A3B (MoE; coding) — desktop only | Poolside AI | OpenMDW License Agreement v1.1 |
AI models are not bundled with NotesXML. They are downloaded separately at the user's request and stored on the user's device. Users are responsible for compliance with each model provider's license terms (links to provider terms are available on each model's Hugging Face model card).
LFM2.5 VL licence — please read if you use NotesXML at work. LFM2.5 VL 1.6B is the recommended default model and is available on the free tier. It is published by Liquid AI under the LFM Open License v1.0, which permits personal use freely, and permits commercial use only by organisations with annual revenue under US $10 million. If your organisation is above that threshold, you will need a separate licence from Liquid AI to use this model commercially, or you can select any other model in the catalog — Ministral 3 3B is also on the free tier, is also vision-capable, and is Apache 2.0 licensed with no revenue condition. NotesXML shows this condition on the model's entry in the AI Models window before you download it. This licence applies to the model only; it places no conditions on NotesXML itself or on the notes you create.
Two models use the OpenMDW License. Nemotron 3.5 Lightning 30B-A3B (NVIDIA) and Laguna XS 2.1 33B-A3B (Poolside AI) are published under the OpenMDW License Agreement, version 1.1 rather than Apache 2.0. OpenMDW is a permissive open model licence covering weights, code and documentation together; it imposes no revenue condition. Both are desktop-only models and neither is on the free tier.
Models retired from the catalog. Llama 3.2 1B Instruct was removed in NotesXML 5.181.4; Llama 3.2 3B Instruct and Phi-4 Mini were removed in the 2026-08-12 catalog revision; LFM2 VL 3B was replaced by its successor LFM2.5 VL 3B in NotesXML 5.188.0. Gemma 4 E2B was removed in 5.181.4 and reinstated in 5.198.33; it appears in the table above. If you downloaded a retired model under an earlier version it remains on your device and continues to work, and it remains governed by the licence listed for its provider.
Llama attribution: Built with Llama. Llama 3.2 models are licensed under the Meta Llama 3.2 Community License, © Meta Platforms, Inc. NotesXML displays this attribution in-app whenever a Llama model is the active selection.
© 2026 IWV Digital Solutions LLC. All rights reserved.