<?xml version="1.0" encoding="UTF-8"?>
<!--
  VoxSpica — PAD 4.0 program description for software catalogs, English version.
  Source of truth: product.json (v0.2.0), downloads.json (exact byte counts),
  landing/src/i18n/en.ts (features). Verify SHA256 against the real file before release.
-->
<PRODUCT version="4.0.0" padversion="4.0" xmlns="http://www.neoworx.com/projects/pad">

  <PRODUCTINFO>
    <NAME>VoxSpica</NAME>
    <PRODUCTTYPE>Application</PRODUCTTYPE>
    <PLATFORM>Windows; Linux</PLATFORM>
    <CATEGORY>Audio, Sound</CATEGORY>
    <SUBCATEGORY>Text</SUBCATEGORY>
    <LANGUAGE>English; Russian; German; French; Spanish; Italian; Simplified Chinese; Ukrainian; Belarusian</LANGUAGE>
    <STATUS>Final</STATUS>
    <DESCRIPTION><![CDATA[
Free offline speech-to-text program for your own computer. Runs on Windows and Linux.

Record from microphone or transcribe an existing audio file — no data ever leaves your machine. Works offline: once the model is downloaded, no network is needed at all. Audio and recognized text never leave the computer — the only network request at startup is a GitHub update check, and it transmits nothing.

Features:
— 33 recognition languages, 9 interface languages (including Ukrainian and Belarusian via the portable build);
— live text as you speak, with pause and stop;
— recognition history in SQLite with search (also from the command line);
— punctuation and capitalization;
— transcribes mp3, m4a, ogg, wav, flac files (ffmpeg needed for mp3/m4a/ogg);
— automatic language detection for audio files;
— choose model size: small (fast, ~0.3 GB RAM) or large (more accurate, ~1.8 GB, loads in 1–2 minutes);
— GUI and command line (VoxSpica.exe mic, file, history --search) sharing settings and history;
— portable build: a single executable in a zip, no Python required, nothing installed.

Two Windows installation options:
— portable build: 44 MB, one executable in a zip, nothing installs;
— per-language installer: ~90 MB, the recognition model is already inside, works offline right after installation. Installers available in English, Russian, German, French, Spanish, Italian and Chinese.

For Linux:
— native .deb for Ubuntu 22.04 / 24.04 and Debian 12+, architecture amd64;
— APT repository: https://alex37529.github.io/voxspica/repo/ (key: https://alex37529.github.io/voxspica/voxspica-archive-keyring.asc).

The program is not code-signed on Windows, so Windows may warn about an unknown publisher and SmartScreen may show a warning. This is the absence of a signature, not malware. The Linux package is signed.
    ]]></DESCRIPTION>
    <SHORTDESCRIPTION><![CDATA[
Offline speech-to-text for Windows and Linux: 33 languages, no cloud, no subscription, no registration. Includes CLI, automatic language detection and small/large model choice.
    ]]></SHORTDESCRIPTION>
    <URL>https://voxspica.com/</URL>
    <HOMEPAGE>https://voxspica.com/</HOMEPAGE>
    <DOWNLOADURL>https://github.com/alex37529/voxspica/releases/download/v0.2.0/VoxSpica-0.2.0-win64.zip</DOWNLOADURL>
    <SUPPORTURL>https://github.com/alex37529/voxspica/issues</SUPPORTURL>
    <DOWNLOADPAGEURL>https://github.com/alex37529/voxspica/releases</DOWNLOADPAGEURL>
    <PUBLISHER>Alexandr Makeenak</PUBLISHER>
    <PUBLISHERURL>https://github.com/alex37529</PUBLISHERURL>
    <PUBLISHEREMAIL>6836808@gmail.com</PUBLISHEREMAIL>
    <AUTHOR>Alexandr Makeenak</AUTHOR>
    <AUTHOREMAIL>6836808@gmail.com</AUTHOREMAIL>
    <AUTHORWEBPAGE>https://github.com/alex37529</AUTHORWEBPAGE>
    <COPYRIGHT>Copyright (c) ALYAKSANDR MAKEENAK. All rights reserved.</COPYRIGHT>
  </PRODUCTINFO>

  <APPLICATIONINFO>
    <APPLICATIONTYPE>Standalone</APPLICATIONTYPE>
    <REQUIREMENTS><![CDATA[
Windows: Windows 10 or 11, 64-bit (x64). Internet needed once — to download the recognition model; after that the program works offline. No graphics card needed, the engine runs on the CPU.

Linux: Ubuntu 22.04 / 24.04 or Debian 12 and newer, x64 (amd64). .deb installs by double-click or apt. APT repository available for automatic updates.
    ]]></REQUIREMENTS>
  </APPLICATIONINFO>

  <LICENSEINFO>
    <LICENSE>Free, no subscription, no ads, no time limit</LICENSE>
    <LICENSETYPE>Freeware</LICENSETYPE>
    <PRICE>0</PRICE>
    <REGISTRATIONREQUIRED>No</REGISTRATIONREQUIRED>
    <HOMEPAGE>https://voxspica.com/</HOMEPAGE>
    <LICENSEURL>https://github.com/alex37529/voxspica/blob/main/LICENSE</LICENSEURL>
    <LICENSEAGREEMENT><![CDATA[
Terms: free use on any number of your own computers.
Redistribution, resale, decompilation and obfuscation of the executable file are prohibited. Source code is closed and not published. The program is provided “as is”, without warranties. Third-party components — VOSK (Apache License 2.0), PySide6/Qt (LGPL v3), PortAudio (MIT), Certifi (MPL 2.0), Python (PSF 2.0) — are distributed under their own licenses.
    ]]></LICENSEAGREEMENT>
  </LICENSEINFO>

  <DOWNLOADINFO>
    <SOFTWARETYPE>Application</SOFTWARETYPE>
    <DESCRIPTIONLANGUAGE>English</DESCRIPTIONLANGUAGE>
    <FILEFORMAT>ZIP</FILEFORMAT>
    <FILESIZE>43766033</FILESIZE>
    <FILEPATH>VoxSpica-0.2.0-win64.zip</FILEPATH>
    <SHA256>9D18E7192C57ADD3563F432F405FF91CADA2FDE1196316B6AB60F191B90D8376</SHA256>
    <DOWNLOADCOUNT>0</DOWNLOADCOUNT>
  </DOWNLOADINFO>

  <SYSTEMREQUIREMENTSINFO>
    <OPERATINGSYSTEM>Windows 10, Windows 11 (x64); Ubuntu 22.04 / 24.04, Debian 12+ (x64)</OPERATINGSYSTEM>
    <PROCESSOR>x86-64, any dual-core from 2 GHz</PROCESSOR>
    <MEMORY><![CDATA[
512 MB for small models. The large model needs about 8 GB RAM: it occupies 1.8 GB and takes 60–90 seconds to load from disk.
    ]]></MEMORY>
    <DISKSPACE><![CDATA[
120 MB for the portable build. The recognition model — from 50 MB (small) to 1.8 GB (large) — is downloaded separately.
    ]]></DISKSPACE>
    <GRAPHICS>Not required, no GPU acceleration is used</GRAPHICS>
  </SYSTEMREQUIREMENTSINFO>

  <CUSTOMINFO>
    <FIELD1><![CDATA[
Seven Windows installers, each with its language model already inside: 93.1 MB (ru), 88.1 MB (en), 93.0 MB (de), 89.3 MB (fr), 86.6 MB (es), 96.4 MB (it), 90.8 MB (zh). Exact byte counts are in the GitHub releases. Links: https://voxspica.com/downloads/VoxSpica-0.2.0-<language code>-setup.exe
    ]]></FIELD1>
    <FIELD2><![CDATA[
Recognition engine — VOSK (Kaldi ASR), computed on CPU. Supports mp3, m4a, ogg, wav, flac; ffmpeg needed separately for mp3/m4a/ogg. Automatic file language detection. Small / large model choice.
    ]]></FIELD2>
    <FIELD3><![CDATA[
Linux: native .deb (amd64) for Ubuntu 22.04/24.04, Debian 12+. APT repository https://alex37529.github.io/voxspica/repo/. Install command: curl -fsSL https://alex37529.github.io/voxspica/voxspica-archive-keyring.asc | sudo gpg --dearmor -o /etc/apt/trusted.gpg.d/voxspica.gpg && echo "deb https://alex37529.github.io/voxspica/repo/ noble main" | sudo tee /etc/apt/sources.list.d/voxspica.list && sudo apt update && sudo apt install voxspica (on Ubuntu 22.04 replace noble with jammy).

CLI: VoxSpica.exe mic, VoxSpica.exe file meeting.mp3, VoxSpica.exe history --search "meeting" — console and window share settings and history database.

Privacy: audio and text are not sent to the internet. The update check contacts api.github.com without transmitting audio or text.
    ]]></FIELD3>
  </CUSTOMINFO>

</PRODUCT>
