Third-party notices

Arc Forge products. Arc Forge products may include, link to, or interoperate with third-party software, models, package repositories, operating-system services, and hosted providers.

Dictate. Dictate is free, local-first desktop dictation software for Windows 11 and Linux. Its speech and speaker models and the open-source components below are included with the app and run on your device.

Provider terms. App stores, GitHub, npm, Hugging Face, model publishers, and package registries apply their own terms, privacy policies, and access rules.

No endorsement. Third-party project names are listed for attribution and transparency. Their maintainers do not endorse Arc Forge or Dictate unless separately stated.

NVIDIA Parakeet TDT 0.6B v2. Dictate’s default speech model, by NVIDIA, licensed under Creative Commons Attribution 4.0 International. Dictate uses the ONNX conversion istupakov/parakeet-tdt-0.6b-v2-onnx of nvidia/parakeet-tdt-0.6b-v2.

onnx-asr and ONNX Runtime. Run the Parakeet model on your device. onnx-asr (github.com/istupakov/onnx-asr) and ONNX Runtime (github.com/microsoft/onnxruntime) are licensed under the MIT License.

faster-whisper and CTranslate2. An alternative local speech engine included with the app, selectable through the config file or command line. faster-whisper (github.com/SYSTRAN/faster-whisper) and CTranslate2 (github.com/OpenNMT/CTranslate2) are licensed under the MIT License.

pyannote.audio. Speaker labelling for the beta meeting feature, licensed under the MIT License. Source and license: github.com/pyannote/pyannote-audio. By default it sends anonymous usage metrics (no audio or text) to pyannote’s telemetry service; see the Dictate Privacy Policy.

pyannote speaker-diarization-community-1. The speaker model for the beta meeting feature, licensed under Creative Commons Attribution 4.0 International. Model page: huggingface.co/pyannote/speaker-diarization-community-1.

Attribution. Parakeet TDT 0.6B v2 by NVIDIA, licensed under CC BY 4.0. pyannote speaker-diarization-community-1 by pyannote.audio, licensed under CC BY 4.0.

Other open-source components. Dictate also includes PyTorch (BSD-style), torchaudio (BSD), PyAV (BSD-3-Clause, with FFmpeg libraries under their own licences), huggingface_hub and tokenizers (Apache-2.0), OpenTelemetry (Apache-2.0), NumPy (BSD-3-Clause), sounddevice and PyYAML (MIT), and pynput (LGPL-3.0). On Linux it also uses python-xlib (LGPL-2.1 or later), evdev (BSD-3-Clause), and the system PortAudio library. Full component list: the Dictate repository.

Included models. Desktop builds include the Parakeet and pyannote model files they need. Dictate downloads other model files from Hugging Face only when you choose a model or setting that is not included.

Hugging Face access. The pyannote model is gated by its publisher. Desktop builds include it; source installs may need a Hugging Face account, an access token, and acceptance of the publisher’s access conditions.

User tokens. If you supply a Hugging Face token for a source install, it is used to access the selected model repository. Keep tokens private and remove them before sharing logs, screenshots, support bundles, or config files.

Updates. Third-party software, model files, licenses, access conditions, and hosting locations may change. Arc Forge may update this notice as packaging choices change.

Creative Commons BY 4.0. The CC BY 4.0 license summary is available at creativecommons.org/licenses/by/4.0.

Hugging Face. Hugging Face model pages can include repository-specific cards, license metadata, and access conditions. Review the selected model page before enabling gated local models.

Microsoft Store. Microsoft Store distribution has separate privacy, certification, and listing requirements. Store-specific disclosures are handled through the Microsoft Store listing.

Contact. Email hello@arcforge.au for third-party notice questions. Last updated 30 September 2026.