The ComfyUI node

Real-time attention heatmaps inside your ComfyUI generation pipelines.

We are opening the node to a small group first. Tell us where to reach you and what you generate, and we will let you know when your slot is ready to run consumer-neuroscience creative validation on Stable Diffusion and FLUX outputs.

Your personal data is processed as described in the Privacy Notice.

Your personal data and communications

Tick the required box to continue. Optional boxes may be left empty.

We open access in small groups. You will hear from us by email when your slot is ready.

Flux.1 dev · txt2img Trusapien · attention Save image 0.9s

Original · AI ad visual

The ad visual as generated: a laundry detergent bottle held out by a sprinter, next to a washing machine.

Drop the generated ad visualor browse files

Predicted attention · first 3s

The same visual with the predicted attention heatmap over it: the hottest area is the sprinter's face, a second warm area sits on the bottle.
Headline 41% Product 28% Logo 3%

Drop the same visual — the heatmap overlays itor browse files

Predicted visual focus score

88 / 100

Above the 74 category median · logo under-seen

Pass
What the node returns Per image · JSON + PNG

→ focus_score: 88 · category_median: 74 · verdict: PASS

headline0.41
product0.28
cta0.17
background0.11
logo0.03

→ note: "Logo is under-seen. Move the mark above the fold or raise its contrast."

→ heatmap.png · 1024×1280 · alpha layer · elements.json

Input
Any IMAGE output
Latency
Under a second
Model
Neurons attention

Sample output · real values come from your own renders

The integration

01

Zero workflow interruption

One node. Plug it between your sampler and Save Image in any pipeline — the image passes through untouched.

02

Predictive heatmaps

See where people look in the first three seconds, per element — headline, product, logo — before the creative ships.

03

Small groups, real setup

We onboard a few teams at a time so each one gets the node tuned to its own pipelines and creative formats.

04

Enterprise security

KVKK and GDPR compliant infrastructure. Images are scored and discarded; nothing is used for training.