VRAM Gated VAE Loader

The VRAM Gated VAE Loader is the smallest member of TrentNodes' gated-loader family, and it does exactly one job: load a VAE only after VidScribe MiniCPM reports that the vision model has unloaded. If you've seen the gated checkpoint loader and thought "why would I need a VAE that waits," the answer is the same one - a caption-then-generate pipeline where the VLM and the diffusion stack would otherwise be fighting over the same VRAM.

In practice you'd usually pair this with the gated checkpoint loader and wire both from the same vram_cleared output. VidScribe finishes, unloads, fires the signal, and both gates open. ComfyUI's scheduling guarantees the loaders don't run until the signal exists, because vram_signal is a required input and a node can't execute before its inputs are produced. That's the whole mechanism, and it's worth being clear about: this is execution-order gating, not a VRAM measurement. It won't stop an OOM caused by a model that's simply too big for your card.

Inputs and outputs

Slim node, slim list:

• vram_signal - connect VidScribe's vram_cleared here.
• vae_name - dropdown of VAEs from ComfyUI/models/vae.
• Output: vae, ready for a VAEDecode or whatever wants a VAE socket.

If you need a VAE that also carries a checkpoint's CLIP and model, reach for the gated checkpoint loader instead - it emits all three. This one is for graphs where the VAE is loaded separately (say, a custom VAE over a standard checkpoint).

When to bother

Honestly? Only in the caption-then-generate graph. If you're not running VidScribe in the same workflow, this node adds a required input for no benefit - a plain VAE Loader is the right tool. When you do have the VLM upstream, though, the gated family turns a fragile "hope the ordering works out" situation into a guaranteed one.

Installing

It ships in TrentNodes under Trent/VLM, so one install covers it. ComfyUI Manager → search "Trent Nodes" → Install is the easy route; if Manager flags the pack (a day-one repo rename left a registry duplicate it still distrusts), clone manually:

cd ComfyUI/custom_nodes
git clone https://github.com/TrentHunter82/TrentNodes
cd TrentNodes
pip install -r requirements.txt

Restart ComfyUI and the whole Trent/ menu - this node included - shows up.