NVIDIA Studio Voice Advanced Settings

Of the three nodes in ComfyUI-NVIDIA-Maxine, this is the one you can skip for your first week. NVIDIA Studio Voice Advanced Settings doesn't touch audio, Docker, or the network. It's a plumbing node in the purest sense - you feed it eight values, it hands you one settings object for NVIDIA Studio Voice Docker Setup to consume. Nothing runs when you execute it.

So why does it exist? The pack runs NVIDIA Studio Voice, a speech-enhancement model from NVIDIA's Maxine family, inside a Docker container on your Windows GPU. The Docker Setup node does that whole job with safe internal defaults - image, container name, target, timeout, even the NGC username are baked in. That setup node deliberately exposes only two inputs to keep beginners from breaking things. This node is the escape hatch for the roughly 10% of the time you need to override one of those defaults.

How it works

Think of it as a config struct on a wire. Its one output, advanced_settings, carries the STUDIO_VOICE_SETUP_SETTINGS type, and you connect it into NVIDIA Studio Voice Docker Setup's optional advanced_settings input. Wire it up and the setup node reads your values instead of its defaults. Leave it unplugged and you get the defaults. It's exactly the "value node fans out one authoritative source" pattern from ComfyUI's plumbing layer, but for a whole bundle of settings at once.

The inputs that matter

Most of the eight inputs you will never touch, and that's fine. The ones worth knowing:

• force_pull (boolean, default false) - the one you're most likely to actually flip. With it off, the setup node reuses an already-downloaded image; turn it on to force a fresh pull and refresh the NIM image if you suspect you're on a stale version.
• target (default 127.0.0.1:8001) - where the local container's gRPC endpoint lives. NVIDIA Studio Voice Enhance calls the NIM at this address.
• wait_timeout_s (default 900) - how long setup waits for the NIM to come up after starting the container.
• ngc_username (default $oauthtoken) - the literal Docker login username NVIDIA's deploy docs use. It is not your NVIDIA account email. Leave it alone.
• image (default nvcr.io/nim/nvidia/studio-voice:latest) and container_name (default studio-voice-nim) - change these only if you have a specific image tag or container naming scheme in mind.

The remaining two, file_size_limit (default ~35 MB) and model_profile (an optional NIM_MODEL_PROFILE override; empty means the NIM picks for you), are genuinely advanced - if you don't know why you'd set them, you don't need them.

How to install it

This node ships inside the pack, so you install the pack once:

cd C:\path\to\ComfyUI\custom_nodes
git clone https://github.com/grrdhdz/ComfyUI-NVIDIA-Maxine.git
cd ComfyUI-NVIDIA-Maxine
C:\path\to\ComfyUI\.venv\Scripts\python.exe -m pip install -r requirements.txt

Restart ComfyUI after installing (or search "ComfyUI-NVIDIA-Maxine" in ComfyUI Manager). The real dependency file is requirements.txt - the README is explicit that there is no requirements.py. Note the runtime floor: this pack targets the V3 custom node API and was validated against ComfyUI 0.21.0, so on an older ComfyUI the nodes won't register.

Common issues

The trap is thinking you need this node at all. If you connect it with the defaults and later change the target or port, NVIDIA Studio Voice Enhance will fail to reach the NIM and throw a "Studio Voice is not ready" error - your override quietly became the source of truth. When in doubt, delete the node and let setup use its defaults. And since force_pull triggers a multi-gigabyte re-download, don't leave it on out of curiosity. It's a settings node with sharp edges that's happiest when ignored.