Instant Load Triple CLIP

ComfyUI ships a single CLIP loader and a dual CLIP loader. It does not ship a triple. Instant Load Triple CLIP is the pack reaching past that: three text encoder dropdowns, one CLIP output, for the handful of recipes that stack three encoders into a single conditioning pipeline.

Be honest about who this is for. The author's own recipe lists suggest dual is still the common case, and triple/quadruple loaders exist for the exotic frontier - experimental multi-encoder models and niche pipelines that fuse more than two text representations. If your workflow doesn't explicitly ask for a triple load, you don't need this node. It's not a "load three models because you can" feature; the CLIPType is hardcoded to the stable-diffusion family here, so it's built for that specific multi-encoder niche rather than a general-purpose combo.

Inputs and output

• clip_name1, clip_name2, clip_name3 - three dropdowns over everything in ComfyUI/models/text_encoders. Order can matter, so match the order your workflow expects.
• CLIP - the fused encoder, fed to CLIPTextEncode like any other CLIP output.

For .safetensors files all three encoders stream through the pack's InstantTensor path: header-read, model construction, then per-tensor copy_() straight into the parameters - so you're not holding three full state dicts in VRAM while they build. That memory saving is the real reason this node exists: three encoders is exactly the load where peak memory gets nasty.

Common issues

The pack rules you've read everywhere else apply here too: CUDA only, the instanttensor pip package must be installed or you get "The instanttensor package is required for CUDA instant loading," and non-safetensors files fall through to ComfyUI's normal loader. One thing that trips people: since this node hardcodes a single CLIP type rather than giving you a recipe dropdown, a file that needs a non-standard encoder treatment won't be handled specially - if the combo fails, check that each encoder individually loads through Instant Load CLIP with the right type first.

How to install

ComfyUI Manager → search ComfyUI-InstantTensorLoaders → install, accepting the dependency install. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/redstonewhite/ComfyUI-InstantTensorLoaders

Restart ComfyUI; the node lives under advanced/loaders/instanttensor. Missing dependency? pip install instanttensor in the venv ComfyUI runs from, then restart. And a note for sharers: the pack is a single-commit, May 2026 release, so anyone opening a workflow that uses this node needs the pack installed or they'll see a missing-node error.