Credits, Foundations & Open-Source Licences
SlyPilot Desktop Helpers stands on a foundation of open-source software, low-level libraries and machine-learning research. We are grateful to the
SlyPilot Desktop Helpers stands on a foundation of open-source software, low-level libraries and machine-learning research. We are grateful to the creators and maintainers of these projects.
What is inside the program and what is downloaded. Only two third-party parts travel inside SlyPilot.exe: the AutoHotkey runtime and VirtualDesktopAccessor.dll. Everything else below (speech engines, models, voices, the built-in language model) is downloaded only if you set that feature up, from the publisher's own site, after a window that lists every file with its size, source and licence, and is checked against a published SHA-256 checksum before it is used.
SlyPilot's free core (the program and its script) is under the MIT License (Copyright (c) 2026 Alan Bedian). The programs of the optional SlyPilot Pro, in the Pro folder, are proprietary and are covered by their own licence terms; the third-party parts they use (for example sherpa-onnx and the voice models for natural voices) are listed below with their own licences.
| Part | Role | Licence | Inside the exe, or downloaded? |
|---|---|---|---|
| AutoHotkey v2 | Language runtime the program is compiled with | GPL v2 | Inside |
| VirtualDesktopAccessor.dll | Virtual desktop control | MIT | Inside |
| whisper.cpp | Speech-to-text engine | MIT | Downloaded (AI speech) |
| OpenAI Whisper models | Speech models | MIT | Downloaded (AI speech) |
| CrispASR | Speech engine (Vulkan GPU, and the CPU build for Parakeet) | MIT | Downloaded (AI speech) |
NVIDIA Parakeet models (GGUF by cstr) | Fast speech models | CC-BY-4.0 | Downloaded (AI speech) |
| sherpa-onnx and ONNX Runtime | Runs the natural voices | Apache-2.0 and MIT | Downloaded (natural voices) |
| Kokoro-82M | Natural voices | Apache-2.0 | Downloaded (natural voices) |
| Piper voices | Natural voices | Per voice dataset (public domain, CC0, CC BY 4.0) | Downloaded (natural voices) |
| Supertonic | Natural voices | Code MIT, voice model OpenRAIL-M | Downloaded (natural voices) |
| espeak-ng pronunciation data | Inside the Kokoro package | GPL-3.0-or-later | Downloaded (natural voices) |
| llama.cpp | Runs the built-in language model | MIT | Downloaded (AI actions, optional) |
| Qwen3.5 models (GGUF by Unsloth) | The built-in language model | Apache-2.0 | Downloaded (AI actions, optional) |
🏛️ Core Foundations & Runtimes
AutoHotkey v2
- Author: Steve Gray (Lexikos) and the AutoHotkey community (github.com/AutoHotkey/AutoHotkey).
- License: GNU General Public License v2 (GPL-2.0).
- Role: The language and runtime that SlyPilot's core engine, Win32 hooks and windows are written in.
SlyPilot.exeis a compiled copy of the AutoHotkey interpreter plus the scriptSlyPilot.ahk, so if you pass the program on, passSlyPilot.ahkwith it (SlyPilot's licence notice asks for this, because the script is the program's source).
VirtualDesktopAccessor.dll
- Author: Jari Pennanen (Ciantic/VirtualDesktopAccessor).
- License: MIT License (Copyright (c) 2015-2023 Jari Otto Oskari Pennanen).
- Role: A 64-bit native library that reaches the Windows virtual desktop interfaces (
IVirtualDesktopManagerInternal). SlyPilot embeds the official release build unchanged; its SHA-256 is shown in the About window so you can compare it with the release.
Windows itself
The Windows speech recognizer used for voice commands, the Windows voices, Windows OCR and the virtual-desktop and window APIs are part of Windows.
🧠 Speech Recognition & Machine Learning (downloaded if you set up AI speech)
whisper.cpp
- Author: Georgi Gerganov and the ggml authors (ggml-org/whisper.cpp).
- License: MIT License.
- Role: A C/C++ port of OpenAI's Whisper speech recognition model, used with NVIDIA CUDA 12 or on the CPU.
OpenAI Whisper Models
- Author: OpenAI (openai/whisper); the converted model files come from huggingface.co/ggerganov/whisper.cpp.
- License: MIT License.
- Role: The speech models (Tiny, Base, Small, Medium, Large v3 Turbo).
CrispASR
- Author: CrispStrobe (CrispStrobe/CrispASR).
- License: MIT License (its other backends and models carry their own licences; SlyPilot uses only the Whisper and Parakeet ones).
- Role: A speech engine that runs Whisper on graphics cards through Vulkan (AMD Radeon, Intel Arc) and runs the Parakeet models on the CPU.
NVIDIA Parakeet Models
- Author: NVIDIA; converted to GGUF format by
cstron Hugging Face (huggingface.co/cstr). - License: Creative Commons Attribution 4.0 International (CC-BY-4.0). Attribution: NVIDIA, converted by cstr.
- Role: Fast first-pass speech models (0.6B English, 0.6B v3 for 25 European languages, and 110M).
🗣️ Neural Speech Synthesis (SlyPilot Pro; downloaded if you set up natural voices)
The 45 voices run in SlyPilot Pro's own helper program, SlyTts.exe, which loads the sherpa-onnx library.
sherpa-onnx
- Author: the k2-fsa / Next-gen Kaldi project (k2-fsa/sherpa-onnx).
- License: Apache License 2.0.
- Role: Runs the voice models and packages them for download.
ONNX Runtime
- Author: Microsoft (microsoft/onnxruntime).
- License: MIT License.
- Role: The neural network runtime that comes with sherpa-onnx.
Kokoro-82M
- Author: hexgrad (hexgrad/Kokoro-82M), packaged by the sherpa-onnx project.
- License: Apache License 2.0.
- Role: An 82-million-parameter voice model. SlyPilot Pro offers its 28 English voices.
- espeak-ng: the Kokoro package contains pronunciation data from espeak-ng (GPL-3.0-or-later). It is downloaded from its publisher at your request and used by the library on your PC; SlyPilot does not redistribute it.
Piper
- Author: the Rhasspy / Open Home Foundation Piper project (OHF-Voice/piper1-gpl), with voices packaged by sherpa-onnx.
- License: each voice follows its own dataset licence. SlyPilot only offers voices whose recordings are public domain, CC0 or CC BY: LJ Speech (public domain), Kathleen (CC0), Joe (CC0), John and Norman (public domain, LibriVox), Cori (public domain, LibriVox) and Alba (CC BY 4.0, University of Edinburgh CSTR, datashare.ed.ac.uk).
- Role: Small, quick voices for short feedback and for PCs with limited memory.
Supertonic
- Author: Supertone Inc. (supertone-inc/supertonic).
- License: Code under MIT License; the voice model under the OpenRAIL-M licence (free use, no harmful or deceptive uses).
- Role: Ten voices in one small download.
✍️ The Built-In Language Model (downloaded if you choose it in AI Actions Setup)
llama.cpp
- Author: Georgi Gerganov and the ggml authors (ggml-org/llama.cpp).
- License: MIT License.
- Role: The small program (
llama-server) that runs the model on your PC, listening only on127.0.0.1.
Qwen3.5 language models
- Author: Alibaba Cloud (the 0.8B and 2B models), with the GGUF files by Unsloth (huggingface.co/unsloth).
- License: Apache License 2.0.
- Role: The small language model that answers AI actions on your PC when you use the built-in option.
Services you may connect yourself (not part of SlyPilot)
AI actions can also talk to Ollama, LM Studio, Ollama's cloud or OpenRouter if you set that up. These are separate products with their own terms and privacy policies; nothing of theirs is bundled, and SlyPilot sends nothing to them until you choose one.
🛠️ Helper Programs Written for SlyPilot
The SlyPilot Pro programs are written in C# and are separate files in the Pro folder next to SlyPilot.exe, not inside it: SlyPro.exe (the Pro service and licence window), SlyTts.exe (natural voices), SlyMeet.exe (meeting recording), SlyRemote.exe (the phone remote's web server). They are proprietary and unlocked by a licence file checked offline. The command-line client SlyCtl.exe sits next to the program. The phone remote's QR-code encoder was written for SlyPilot after the public description of the QR standard and Project Nayuki's reference implementation.
📜 Full License Notices
The full text of the licences of the parts that travel with SlyPilot is in THIRD_PARTY_LICENSES.md in the distribution, and inside the program under Tray menu → About / Diagnostics... → Licences.