lens / turbo / createImage

lens / turbo / createImage is the throttle-limit version of the Lens engine, and it holds the distinction of being the most aggressive speed/quality trade in the whole Civitai pack: 4 steps at cfg 1.0. Four. That's a distilled-model number, the same territory as a turbo LoRA or a schnell-class checkpoint. If the normal Lens node is a solid generalist at 20 steps, this is its espresso shot - fast, cheap, and not remotely meant to be your final render.

As with its normal sibling, "Lens" is a barely-documented engine name that shows up only in Civitai's orchestration spec and this generated node menu. What's knowable is the node shape: hosted text-to-image, billed in Buzz, prompt in, native IMAGE tensor out. The turbo variant's only reason to exist is throughput - and with quantity up to 12 per call, the intended use is obvious. Wire it into a batch loop, generate a wall of thumbnails, pick a direction, then send the winner to a proper model. Using a 4-step node for finals is how you end up with muddy midjourney-esque mush and a confused wallet.

The inputs that matter

prompt required, everything else has a default.

• steps (4) / cfg_scale (1) - leave them alone. The moment you start cranking cfg on a 4-step turbo model you get harsh, doubled artifacts; that's the model telling you it was distilled, not that it's broken.
• sampler / scheduler - exposed but pointless to fiddle with for a draft node; the platform defaults are fine.
• width / height - 1024 default, up to 2048. The turbo pitch collapses at the top end; drafts at 1024 are the honest use.
• quantity - up to 12 in one billed job. This is the node's whole reason to exist.
• diffusion_model / loras - CIVITAI_AIR / CIVITAI_LORAS sockets for the Civitai Model/LoRA Selector nodes, if you want to steer what the Lens pipeline runs.
• negative_prompt, seed, output_format, image_metadata - standard.

Outputs are the pack standard: images (IMAGE), errors, workflow_id, raw_json, with the job id and cost shown on the node.

Install and auth

One install for the whole pack - ComfyUI Manager → Civitai Comfy Nodes, or:

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

Authenticate with a Civitai account that has Buzz: export CIVITAI_API_TOKEN=..., a Civitai Auth node, or the sidebar sign-in.

The bottom line

If you're comparing the two Lens nodes, this one wins on cost-per-draft and loses on everything else, by design. It's the node to reach for when you want to fill a canvas with cheap options, not when you want one good image. And because Lens itself is so thinly documented, treat it as experimental: run a single test draft first, and don't build a production workflow around an engine nobody can tell you much about. If you want a fast, well-documented Krea 2 alternative in the same pack, the Krea 2 turbo node is the safer bet.