Nodes/ComfyUI Assistant Node/PVL KONTEXT MAX MULTI (fal.ai)
ComfyUI Node

PVL KONTEXT MAX MULTI (fal.ai)

Merge two or more images with Kontext Max — one instruction, up to four references

By pvlprk·Created about a year ago·Updated 8 months ago· 1
PVL KONTEXT MAX MULTI (fal.ai)
  • image_1
  • image_2
  • image_3
  • image_4
  • image
prompt
aspect_ratio1:1
guidance_scale3.50
num_images1
safety_tolerance2
output_formatpng
sync_modefalse
seed0
api_key

Kontext Max is the flagship end of the Kontext line, and the "Multi" variant is the one that takes several images at once - merge your character into someone else's scene, combine a product photo with a background, hold two reference images and blend them under one instruction. This node wraps fal's flux-pro/kontext/max/multi endpoint with a minimum of two and up to four image inputs, plus a prompt telling the model what to do with them. If the pack's single-image Kontext nodes are the everyday editors, this is the compositor.

It's part of the ComfyUI Assistant Node pack from pvlprk, and it's one of the few nodes in the pack with its own api_key input - handy if you run different fal accounts per workflow.

How it works

The required inputs are image_1 and image_2; image_3 and image_4 are optional, and the node needs at least two to bother calling the API. All connected images are base64-encoded and sent as a list with prompt, and the model interprets the instruction across all of them - "put the person from image 1 into the setting of image 2" is the genre. The controls are Pro-tier sparse: aspect_ratio (1:1 default, with the usual 21:9 → 9:21 set plus a None option), guidance_scale (3.5), num_images, safety_tolerance (1–6), output_format, sync_mode, and seed.

Two details worth knowing. The api_key field, if filled, sets FAL_KEY in the environment for the call - a per-workflow override instead of a global env var. And the node's error path returns a blank image rather than raising, so a botched multi-image call can look like a successful one.

The inputs that matter

  • image_1 / image_2 - the two required references.
  • image_3 / image_4 - optional extra references.
  • prompt - the instruction tying them together.
  • aspect_ratio - output shape.
  • api_key - optional per-node fal key override.

Output: one IMAGE.

Installing it

Part of pvlprk/comfyui-pvl-api-nodes. Install via ComfyUI Manager (search "ComfyUI Assistant Node") or:

cd ComfyUI/custom_nodes
git clone https://github.com/pvlprk/comfyui-pvl-api-nodes

Set FAL_KEY (or use the node's api_key field), restart, add credits.

Where people get burned

The blank-on-error behavior is the trap: if you get an image that's clearly not what you asked for, don't assume the model misread you - check the console for the real error. Also, "Multi" means the request carries more image data and runs a bigger model, so per-call cost and latency are the highest in the Kontext family; iterate on one image first. And remember Kontext Max is API-only, so none of this runs locally no matter what - the node is the entire setup.

CategoryPVL_tools_FAL

Inputs (13)

NameTypeDefaultDescription
promptSTRING
image_1IMAGE
image_2IMAGE
image_3optIMAGE
image_4optIMAGE
aspect_ratiooptCOMBO1:110 options: , 21:9, 16:9, 4:3, 3:2, 1:1, +4
guidance_scaleoptFLOAT3.501–20
num_imagesoptINT11–4
safety_toleranceoptCOMBO26 options: 1, 2, 3, 4, 5, 6
output_formatoptCOMBOpng2 options: jpeg, png
sync_modeoptBOOLEANfalse
seedoptINT00–4294967295
api_keyoptSTRING

Outputs (1)

NameTypeDescription
imageIMAGE