ComfyUI Node

FLUX.1 Kontext [max]

The API-tier instruction editor

By Runware·Created 2 years ago·Updated about a month ago· 140
FLUX.1 Kontext [max]
  • referenceImages
  • advancedFeatures.watermark.image
  • image
positivePrompt
size1K (1:1)
seed0
numberResults1
advancedFeaturesfalse
advancedFeatures.watermark.bgColor
advancedFeatures.watermark.displayPosition(default)
advancedFeatures.watermark.fontColor
advancedFeatures.watermark.opacityfalse
advancedFeatures.watermark.opacity_value0.10
advancedFeatures.watermark.text
providerSettings.bfl.promptUpsamplingfalse
safetyfalse
safety.checkContentfalse
providerSettings.bfl.safetyTolerance2
ttlfalse
ttl_value60
outputFormatJPG
outputQuality95

FLUX.1 Kontext [max] is the closed, API-only version of the model that turned instruction-based editing into a commodity. Same family as the Kontext dev you might already know from local workflows - feed it reference image(s) and a sentence, get back an edit that keeps identity - but this one runs on Black Forest Labs' servers with the max-tier quality and none of the license headaches that the dev checkpoint shipped with. You trade away local control for "it just works" at a higher quality ceiling.

If the pack's RunwareArch_flux_1_kontext_dev node is the "bring your own checkpoint" version, this is the turnkey one: the model AIR (bfl:4@1) is baked in and you never think about weights again.

How it works

Kontext is architecturally interesting because it consumes reference images in-context, not through adapter embeddings - which is exactly why character identity survives an edit. This node is a taskType: imageInference REST call; your image goes up as a data URI, the instruction rides in the prompt, and the result comes back as an IMAGE tensor.

Inputs that matter

  • positivePrompt - required. This is the instruction: "swap the background to a beach at sunset," "change her coat to a red puffer."
  • referenceImages - the source image(s) to edit or reference, wired from Load Image. This is the whole point of the model, so get comfortable feeding it a proper batch.
  • width / height - default 1024; no tight range in the schema, so match your reference's aspect roughly.
  • providerSettings.bfl.promptUpsampling - the auto-expander; nice for terse instructions.
  • providerSettings.bfl.safetyTolerance - 0–5, default 2, the BFL moderation dial.
  • seed / numberResults - reproducibility and variants.

Install

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

Or ComfyUI Manager → search Runware → Install → restart. API key in ComfyUI Settings → Runware API key or RUNWARE_API_KEY. Prepaid top-up; per-run cost on the title bar.

Gotchas

  • The Kontext family's party trick - clean object/watermark removal - is also its licensing sore spot. The dev tier's license update about circumventing content restrictions pushed a lot of users out; max sidesteps that by being an API you're renting, but you still answer to BFL's moderation.
  • Results depend heavily on the reference image quality. Feed it a blurry reference and expect a blurry edit.
  • No steps or CFG widgets - provider-fixed. If you need dials, the Kontext dev arch node is the knob-rich alternative.

For character-consistent editing without the 12B-model footprint or the license ambiguity, this is the cleanest path. It's the API version of the model the community called "shockingly effective."

CategoryRunware/Image/bfl

Inputs (21)

NameTypeDefaultDescription
positivePromptSTRINGText prompt describing elements to include in the generated output.
referenceImagesoptIMAGE
advancedFeatures.watermark.imageoptIMAGE
sizeoptCOMBO1K (1:1)Output size. Only one of resolution or dimensions is sent.
seedoptINT00–4294967295Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range.
numberResultsoptINT11–20Number of results to generate. Each result uses a different seed, producing variations of the same parameters.
advancedFeaturesoptBOOLEANfalseEnable to set advancedFeatures. Off uses the model's default.
advancedFeatures.watermark.bgColoroptSTRINGBackground color in hex format.
advancedFeatures.watermark.displayPositionoptCOMBO(default)Watermark position.
advancedFeatures.watermark.fontColoroptSTRINGText color in hex format.
advancedFeatures.watermark.opacityoptBOOLEANfalseEnable to set advancedFeatures.watermark.opacity. Off uses the model's default.
advancedFeatures.watermark.opacity_valueoptFLOAT0.100.1–1Watermark opacity.
advancedFeatures.watermark.textoptSTRINGWatermark text.
providerSettings.bfl.promptUpsamplingoptBOOLEANfalseEnable automatic enhancement and expansion of the input prompt.
safetyoptBOOLEANfalseEnable to set safety. Off uses the model's default.
safety.checkContentoptBOOLEANfalseEnable or disable content safety checking.
providerSettings.bfl.safetyToleranceoptINT20–5Tolerance level for content moderation. Lower values are stricter; higher values are more permissive.
ttloptBOOLEANfalseEnable to set ttl. Off uses the model's default.
ttl_valueoptINT60Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`.
outputFormatoptCOMBOJPGFile format for the generated image.
outputQualityoptINT9520–99Compression quality of the output. Higher values preserve quality but increase file size.

Outputs (1)

NameTypeDescription
imageIMAGE