Juggernaut Lightning Flux

Lightning was always Juggernaut's speed lane. When the SDXL Juggernaut line released its Lightning variant in early 2024, the pitch was 4–6 step generation - the distilled, fast version of the realism model everyone else was already using. This node is that idea transplanted to Flux: a distilled checkpoint (AIR rundiffusion:110@101) hosted on Runware, built for quick renders that still carry the Juggernaut photography DNA.

Reach for it when speed matters more than the last 10% of fidelity - iterating on prompts, batch-testing compositions, or generating on hardware that can't hold a Flux checkpoint anyway (which is the whole point of the pack: this runs in the cloud, not on your machine). It's the "sketch fast, render final" node. When you've nailed the prompt, swap to the Base Flux node for the careful pass.

How it works

Same node contract as the Base Flux sibling, minus one feature. positivePrompt, width, height (128–2048, step 64) in; imageInference request over REST; native image (IMAGE) out. The difference is the checkpoint and its profile: distilled models want fewer steps, and the node's steps gate exists for exactly that reason - flip it on and you'll find single-digit steps get you most of the way there.

Inputs that matter:

• negativePrompt - excluded from generation when guidance is off (CFGScale ≤ 1).
• seed - set for reproducible runs, 0 for random.
• steps / steps_value - off (model default) until you enable it; for a Lightning model this is the lever that defines "fast."
• CFGScale / CFGScale_value and strength / strength_value - the same gate-toggle pattern; the model has usage rules, so defaults stay unset.
• numberResults - up to 20 variations.

Then the builder sockets: seedImage, maskImage, lora, controlNet, ipAdapters, ultralytics, outpaint, acceleratorOptions. Notice there's no pulid socket here that the Base node has - the Lightning checkpoint doesn't expose it.

Output: image (IMAGE); advanced_json takes hiresFix.

Installing

One install for the pack: ComfyUI Manager → Custom Nodes Manager → search Runware → install → restart, or:

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

No model downloads, light deps (runware-sdk, Pillow). API key from the Runware dashboard.

Gotchas

The name "Lightning" is doing real work here but don't read too much into it - you're still paying per render, and "fast" means fewer steps, not free. Check the per-run cost on the title bar. The gate-toggle inputs are off by default by design (usage rules); if your output surprises you, verify you actually enabled the ones you think you set. And know your lane: Lightning for iteration and drafts, Base Flux for the final image. That split is the honest way to use the two Juggernaut nodes together.