lens / normal / createImage

Let's be straight with you: "Lens" is the least documented engine in this pack, and this is the node for it. civitai-comfy-nodes is generated straight from Civitai's orchestration API spec, and the spec says there's an image engine called lens reachable through the comfy ecosystem, with a normal and a turbo variant. That's roughly the entire public documentation. There's no Civitai model card that names it, no open-weights repo to point at, and the community discussion barely registers. What we can tell you with confidence is the shape of the node: it's a hosted text-to-image call, billed in Buzz, that returns a native IMAGE tensor into your graph - the standard cloud-recipe pattern this pack runs everywhere.

The normal variant defaults to 20 steps at cfg 4.0 at 1024×1024, which puts it in the "solid generalist" band of the menu - same defaults as Krea 2 base, for reference. The turbo sibling is 4 steps at cfg 1.0, which is about as aggressive a speed/quality trade as this pack makes. If you're deciding between the two, normal is the one to try first; the model behaves like a conventional CFG-based diffusion model and responds to sampler choices and cfg tuning the way you'd expect.

The inputs that matter

prompt is the only required input. Beyond that:

• negative_prompt - supported and worth using; the node treats it like a standard SD-style negative.
• sampler / scheduler - full ComfyUI menus (euler, euler_ancestral, heun, dpm_2… plus the scheduler list). The platform defaults are fine; these exist for people who know what they want.
• steps (20) / cfg_scale (4) - the normal defaults.
• width / height - 1024 default, up to 2048.
• quantity - up to 12 per call, the same generous batch knob as the Krea nodes.
• diffusion_model / loras - CIVITAI_AIR and CIVITAI_LORAS sockets, filled from the Civitai Model/LoRA Selector nodes. This is the interesting bit: the Lens engine takes a model socket, so "Lens" is a pipeline, not one frozen checkpoint - the selector is how you choose what actually runs.
• seed, output_format, image_metadata - standard.

Outputs: images (IMAGE), errors, workflow_id, raw_json. The node shows the job id and cost inline.

Install and auth

Shared with the whole pack. ComfyUI Manager → Civitai Comfy Nodes → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt

Then authenticate (Civitai account with Buzz): export CIVITAI_API_TOKEN=..., a Civitai Auth node, or the sidebar connect panel.

Honest advice

If you found this page because you googled "Lens ComfyUI," the odds are you want something else - a lens simulator, an upscaler, Google's thing - and this won't be it. If you're genuinely poking through the Civitai menu, treat this node as an experiment: the model socket means it's worth wiring up a Model Selector and seeing what the ecosystem actually runs behind the name. And keep expectations measured - for a model this undocumented, the safest move is to test one cheap draft before you commit a workflow to it. Flagging this one as a genuine knowledge gap on our side: nobody in the modidex corpus has anything useful to say about it yet.