☁️BizyAir Flux Kontext API

Kontext is Black Forest Labs' instruction-based image editor: hand it an image and a sentence - "remove the logo," "make the jacket red," "put her on a beach" - and it edits in place instead of regenerating from scratch. That in-context approach is what made it the go-to for consistent character edits and, famously, watermark removal. This node calls the hosted Pro/Max tiers through SiliconFlow rather than running the open Kontext Dev weights locally.

How it works

Feed it a prompt and, optionally, an image. With an image attached, the prompt acts as an edit instruction against it; omit the image entirely and you're effectively using Kontext's API as a straight text-to-image generator on the same Pro/Max endpoint. That optional-image behavior is the main thing to know about this node's shape.

Routing through the API tiers rather than the open Dev weights matters for one specific reason: Kontext Dev's license carries an anti-circumvention clause that turned NSFW Kontext LoRAs into takedown targets across CivitAI and HuggingFace through mid-to-late 2025. That's a Dev-weights problem specifically - nobody is downloading or fine-tuning anything through this node, you're paying per call for BFL's hosted model, so BizyAir/SiliconFlow's own server-side content policy applies instead. Don't assume going through an API relay makes content restrictions looser; if anything it's the opposite, since there's no local weights file to route around.

The inputs and outputs that matter

• prompt - required, multiline. The edit instruction, or the generation prompt if you're skipping the image.
• model (default flux-kontext-pro, or flux-kontext-max) - Max is BFL's tier with better prompt adherence and typography, at presumably higher cost/latency than Pro.
• aspect_ratio (default 16:9; choices 21:9, 16:9, 4:3, 1:1, 3:4, 9:16).
• seed (0-2,147,483,647).
• image (optional, IMAGE) - the image to edit. Leave it out for text-to-image.

Outputs: image, plus bizyair_model_name (mapping both model choices to a flux-kontext billing bucket).

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

Full-body edits go wrong around the head and hair. This is a known Kontext-family limitation, not something specific to BizyAir's relay: once a full body fits into frame at the model's working resolution, heads and hair start showing compression-like artifacts and proportions can skew. Crop tighter on the subject before editing, or plan on a touch-up pass afterward for full-body shots.

Not sure whether to include an image. Include it whenever you want an edit against something specific; omit it if you just want Kontext's Pro/Max tier as a plain generator - both are valid uses of the same node.

Prompt gets refused or altered. That's the API's own content moderation, upstream of BizyAir - there's no local weights file here to work around it with.