☁️BizyAir Kling 2 Image To Image
BizyAir Kling 2 Image To Image (BizyAir_Kling_2_I2I_API) — ComfyUI Node
- image
- images
- bizyair_model_name
What it is
Kling isn't only a video model - this node calls into Kling 2's image-to-image endpoint, taking a source image plus a prompt and reworking it accordingly. It's the same closed, subscription-grade model people compare open video generators against, here doing still-image editing instead. As with every Kling node in this pack, the value is access without a separate account: BizyAir wraps Kuaishou's API and bills it against your existing credit balance.
The inputs that matter
- image - your source picture. The schema's own tooltip is refreshingly direct about it: this is the input image.
- prompt - describes the edit or transformation you want.
- model_name - locked to
kling-v2. - aspect_ratio - a genuinely wide set of options: 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9.
- resolution - 1K, the only tier this node offers (Kling's text-to-image node, elsewhere in this pack, also offers 2K - worth knowing if resolution matters more than the image-to-image workflow).
- variants - how many candidate outputs to generate per call, 1 to 9. Generating a few variants at once and picking the best is often more efficient than re-running one at a time when you're not sure exactly how strong an edit you want.
- image_fidelity - 0 to 1, default 0.5. This governs how closely the result sticks to your source image versus how much license the prompt gets to reshape it. Push it toward 1 to keep the original mostly intact with a light edit; pull it toward 0 to let the prompt take over more aggressively.
Outputs: images and bizyair_model_name, the latter reporting which backend model version served the request.
How to install it
ComfyUI Manager: search "BizyAir," install, restart. Or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAir.git
A BizyAir API key is required - there's no local fallback for Kling, closed model, cloud only.
Common issues
Edit is too aggressive or barely changes anything. image_fidelity is the first dial to reach for either way - raise it if the source is getting overwritten more than you wanted, lower it if the edit feels too timid relative to your prompt.
Wasting variants on a prompt you're not confident about yet. Since each of the 1-9 variants counts against your usage, it's worth iterating on the prompt itself at variants: 1 first, then bumping the count up once you're happy with the direction and just want to pick the best execution of it.
Aspect ratio doesn't match your source image and results look cropped or stretched oddly. With eight aspect ratio options available, match it to (or deliberately choose a crop from) your source rather than leaving it on a default that doesn't fit your input.
Job fails or times out. Check your API key and account balance before assuming the prompt or settings are the problem - this is a network call to a third-party service routed through BizyAir, and auth/billing issues are the most common source of an opaque failure.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | 输入图片 | |
| prompt | STRING | — | |
| model_name | COMBO | kling-v2 | 1 options: kling-v2 |
| aspect_ratio | COMBO | 16:9 | 8 options: 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, +2 |
| resolution | COMBO | 1K | 1 options: 1K |
| variants | INT | 11–9 | — |
| image_fidelity | FLOAT | 0.500–1 | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| bizyair_model_name | {"kling-v2": "kling-v2"} | — |