Neural Companion is currently a Windows-first Python 3.11 desktop app.
The recommended public install path is the unified installer:
Graphical installer:
INSTALL_NEURAL_COMPANION.bat
If you prefer launching it from a terminal:
py install_neural_companion_gui.pyThe graphical installer detects Python 3.11, offers a Windows Python download
link when it is missing, can install bundled FFmpeg tools when ffmpeg or
ffprobe are not available, and includes links for Discord setup help plus the
Hugging Face token/model-term pages used by PocketTTS.
For RTX 50 / Blackwell GPUs, the installer can auto-select or manually force the CUDA 12.8 PyTorch stack used by the main runtime and MuseTalk.
Command-line installer:
py install_neural_companion.py --main --non-interactiveFor a fuller local setup:
py install_neural_companion.py --allTo install the default Echo and Eon MuseTalk avatar packs from the separate avatar-pack release:
py install_neural_companion.py --avatar-packs --non-interactiveThe graphical installer selects both default avatar packs by default. The
command-line --all target installs the main, MuseTalk, and PocketTTS runtimes;
combine it with --avatar-packs when you also want the avatar packs.
If Python 3.11 is not your default Python, pass the interpreter explicitly:
py install_neural_companion.py --python-exe "C:\Path\To\Python311\python.exe"The installer also tries to find Python 3.11 automatically through the Windows
Python launcher, PATH, and common install folders. The graphical installer shows
the detected interpreter and lets you browse to another python.exe.
requirements.txtis the small baseline requirement list.requirements.companion.txtis the broader main-app runtime set used by the installer.requirements.musetalk.txtis the separate MuseTalk runtime set.
Most users should prefer the installer over manually installing requirement files. The installer keeps incompatible runtime stacks separated where needed.
MuseTalk:
py install_neural_companion.py --musetalk --non-interactivePocketTTS:
py install_neural_companion.py --pockettts --non-interactiveThe main app can run without MuseTalk, PocketTTS, VaM, VSeeFace, or API
providers. Start with None avatar mode and one working chat provider first.
- Start LM Studio or configure an API chat provider.
- Run
run_neural_companion.batorpy qt_app.py. - Select a chat provider and model.
- Select
Noneas Avatar Engine. - Select a TTS backend.
- Press
Initialize System. - Verify chat and speech before enabling heavier avatar or vision features.
The public repo does not ship voice samples, avatar packs, model weights, or generated images.
- Put your own voice references in
voices/. - Put MuseTalk avatar packs in
avatar_packs/<pack_id>/.
Demo avatar packs live in:
https://github.com/Rakile/NeuralCompanion-AvatarPacks
See docs/avatar_packs.md and docs/release_asset_policy.md for the asset
policy.