Skip to content

System Requirements

UltiVoice runs the full AI pipeline locally. The models are large — review these requirements before installing.

Requirement
OSWindows 10 (64-bit, version 1903 or later) or Windows 11
Architecturex86-64 (ARM not supported in 3.x)
ComponentMinimumRecommended
CPU4-core x86-64, 2.5 GHz8-core, 3.5 GHz+
RAM8 GB16 GB or more
GPUNone (CPU-only mode)NVIDIA GPU with 6 GB VRAM (CUDA 11.8+)
Disk (free space)20 GB60 GB+
InternetRequired for activation only
ItemApproximate size
UltiVoice installer + app~200 MB
Whisper large-v3 (ASR model)~3 GB
Qwen2.5-7B GGUF (translation)~5 GB
F5-TTS-Vi voice model~1 GB
Demucs audio separation~1 GB
Working space (temp files)5–20 GB depending on video length

Models are downloaded on first use and stored in %APPDATA%\com.btgroup.ultivoice\models. You can point to an existing download location in Settings → Models.

UltiVoice is offline-first. Internet access is only needed for:

  • Initial license activation — a one-time call to the License Authority.
  • Model downloads — first-time download of AI models (several GB; can be done on a separate machine and copied over).
  • Subscription renewal checks — periodic background ping; the app continues working offline if the server is unreachable (grace period applies — see Offline Use & Renewal).

UltiVoice bundles all required runtimes. You do not need to install Python, FFmpeg, or CUDA separately. The installer places everything under the app directory.

Bundled componentVersion
Python runtime (sidecar)3.12
FFmpeg6.x
CUDA runtime (if GPU present)11.8

Some antivirus tools flag PyInstaller-frozen executables. If the installer or sidecar is blocked:

  1. Add the UltiVoice install directory to your AV exclusion list.
  2. Allow outbound HTTPS (port 443) for ultivoice.com (license activation) and model download hosts.

See Troubleshooting → Install fails if your AV quarantines the app.