☁️BizyAir Generate Photorealistic Images
A fast, cloud-hosted distilled generator
- IMAGE
The class name - GenerateLightningImage - is the tell here. "Lightning" in this ecosystem almost always means a step-distilled model tuned for speed, the same lineage as SDXL-Lightning and similar fast checkpoints, running on SiliconFlow's GPUs rather than yours. This node is a self-contained text-to-image generator: no separate checkpoint, sampler, or scheduler nodes to wire up, just a prompt and a handful of dials.
How it works
Distilled, "lightning"-style models bake an approximation of guidance into their weights the same way Flux does, which is why this node's default cfg sits at 1.5 rather than the 5-9 range a normal SDXL checkpoint would want. Pushing cfg up on a model like this tends to burn and oversaturate the image instead of sharpening prompt-following - the same failure mode guidance-distilled and step-distilled models share across the board.
The inputs and outputs that matter
prompt(default"a dog") - multiline, required.seed- an unusually large range here, 0 up to roughly 1.8×10¹⁹, well beyond the 2³¹ ceiling most nodes cap seeds at.width/height(default 1024, range 16-1024, step 8) - this is a fixed-resolution fast generator; it doesn't scale up to 2K the way the API-tier nodes in this pack can.cfg(default 1.5, 0-10, step 0.1) - keep it low. This behaves like a distilled model's guidance value, not a classic SDXL CFG slider.batch_size(default 1, 1-4) - how many images per call. Since this runs on BizyAir's infrastructure rather than your own GPU, the real constraint on batching is your credit spend and BizyAir's rate limits, not local VRAM.
Output: a single IMAGE (the batch, if batch_size > 1).
There's no negative-prompt input on this node at all - nothing in the schema for it. If you need to steer away from something, state it as a positive constraint in prompt rather than expecting a negative field to show up, which lines up with how low-CFG, guidance-baked-in models generally behave anyway: they mostly can't use a real negative pass.
How to install it
ComfyUI Manager: search BizyAir, install, restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAir
then restart. This is a paid cloud call - set your API key on first use per the README and expect it to draw from your BizyAir credits per generation.
Common issues & troubleshooting
Images look burned or oversaturated. Check cfg first - anything much above the 1.5 default is the likely cause on a distilled-style model like this one.
Wanting a negative prompt. There isn't one. Rephrase whatever you were going to put there as a positive statement in prompt instead - "clean plain background" beats "no clutter."
Wanting resolution above 1024. Not available on this node - it's capped at 1024×1024. Reach for one of the pack's other API generators (Flux 2, for instance) if you need larger output.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | a dog | — |
| seed | INT | 10–18446744073709550000 | — |
| width | INT | 102416–1024 | — |
| height | INT | 102416–1024 | — |
| cfg | FLOAT | 1.50–10 | — |
| batch_size | INT | 11–4 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |