Flux Kontext Dev LoRA
Flux Kontext Dev editing in ComfyUI, LoRAs included
- images
- api_config
- image
- request_id
- urls
Flux Kontext was Black Forest Labs' instruction-based editing model: you give it an image and a sentence - "change the jacket to leather," "remove the watermark," "put her in a rainstorm" - and it returns the edited image while keeping the subject intact. Kontext Dev, the 12B open-weights tier, is one of the few genuinely good editors you can host yourself. This BizyAirPlus node gives you Kontext Dev as a cloud call, with a LoRA selector attached, which is where it gets more interesting than a plain remote editor.
What it is
A "Self-Hosted" category node: BizyAir runs the open-weight Kontext Dev model on its own infra and you call it over HTTP. images + prompt in, edited image out. The twist is the loras input - this isn't just "Kontext Dev, the vanilla model," it's Kontext Dev with BizyAir's model selector for LoRAs, so you can drop a style or character LoRA onto the edit. That's the difference between an editing node and an editing setup.
How it works
Kontext's architecture is in-context editing: the reference image is concatenated into the model's context rather than injected as an adapter embedding, which is why it holds identity so well across successive edits. The BizyAir node handles upload, inference, and return - you just feed the image and the instruction. The loras field pulls from BizyAir's hosted LoRA library, so the strength/style of the edit is tunable in a way the stock model doesn't give you.
The inputs that matter
- images - required. The input picture to edit.
- prompt - required. The instruction, in plain language: what to change, what to keep.
- loras - optional. Choose a LoRA from BizyAir's selector to steer style or identity.
- seed - reproducibility,
-1for random.
api_config and skip_error are the pack standard.
Outputs
image wires into a save/preview. request_id is the cloud job ID. urls gives direct links to the edited files.
Installing
ComfyUI Manager → search "BizyAirPlus", or:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart, toggle the BizyAirPlus button ON, enter your API key from bizyair.ai.
Where people get burned
Know what Kontext Dev is and isn't. It's the community's legendary watermark remover and a strong character-consistency editor - but full-body shots degrade anatomy (heads and hair artifact when a full figure fits the frame), and BFL's Kontext-era license clause about circumventing content restrictions made NSFW LoRAs a license violation, with real enforcement against platforms. So: great for legitimate editing, not a tool to get cute with. Also, instruction editing is iterative by nature - budget for a few prompt passes per image, since each is a metered cloud call.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | Input Picture | |
| prompt | STRING | Prompts | |
| lorasopt | STRING | Choose Lora | |
| seedopt | INT | 0-1–2147483647 | Seed |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| request_id | STRING | — |
| urls | STRING | — |