Computational neuroscience

The human brain. Reimagined.

We’re developing a computational model of how the human brain responds to any kind of environmental stimulus. Its predictions offer a window into human attention and experience.

Explore applications How it works
01 / Experience

Experience beginswith a stimulus.

Images, video, sound, and text form the input to the model. This example follows a film from stimulus to predicted response.

01 Stimulus

Video example · response over time

Loading visualization

Kovex / Prediction workflow
Model outputs

Predicted brain activity
and visual attention.

Our models estimate brain responses to visual, audio, and text stimuli as EEG signals over time. For visual content, gaze predictions add spatial detail, indicating which regions are likely to receive attention.

01 / Brain activity

Predicted EEG

Predicted patterns of electrical brain activity over time provide a basis for estimating attention, arousal, memory encoding, and other measures relevant to the task.

Illustrative EEG · Not a recording from a viewer
02 / Visual attention

Predicted gaze

Gaze heat maps estimate where people are likely to look within a scene, revealing how objects, interface elements, and messages may compete for visual attention.

Loading model-generated gaze…
How it works

From stimulus
to response metrics.

Kovex takes images, video, sound, or text as input and estimates the brain’s response. Predicted signals are translated into measures that describe how that response varies over time and across different stimuli.

01 / →

Sensory input

Images, video, sound, and text provide the stimulus. Different versions of the same material allow comparisons between designs, edits, or experimental conditions.

02 / →

Response prediction

Kovex models the response over time as predicted EEG signals, with gaze predictions for visual content.

03 / Insight

Response measures

Measures derived from the predicted signals describe aspects of attention, arousal, and memory encoding. Their variation can reveal differences between moments, content variants, or experimental conditions.

Interpreting the metrics

Attention, arousal, memory encoding, and cognitive load describe different aspects of a response. Each measure needs to be interpreted and validated for the stimulus, population, and application under study.

Applications

Applications in design
and research.

Every field already tests with people: pre-tests, playtests, recordings, participant studies. Kovex runs one step earlier, so the versions that reach people are the ones worth testing.

Frame from the adidas Football Anthem Film: looking up through a goal net at players celebratingSource film · adidas

01Media & advertisingBefore pre-tests

Is the brand on screen when attention peaks?

Input
Rough cuts, the hero film and its 15s and 6s cutdowns, or two edits of the opening seconds.
Output
Where viewers look frame by frame, and when attention and memory encoding rise or drop.
Use
Pick which edits go to a paid pre-test, and move the brand or product into the moments that hold attention.
A · Football Anthem FilmB · You Got ThisPredicted attention · 0–88s
  • Gaze
  • Attention
  • Memory encoding
Case study below ↓
First-person view down a space station corridor toward an illuminated objective doorway, with the game HUD visibleConcept still · generated

02GamesBefore playtests

Do players notice the cue before they need it?

Input
Gameplay captured from two builds: waypoint styles, HUD layouts, or an enemy’s attack wind-up.
Output
Whether predicted gaze reaches the cue or stays on the HUD and effects, and how attention shifts through the sequence.
Use
Fix readability problems before a playtest, so player time goes to the design questions only people can answer.
  • Gaze
  • Attention
Grating stimulus · rendered

03Neuroscience researchBefore recording

Which stimuli will actually separate your conditions?

Input
A candidate stimulus set for each condition: images, clips, or timing and orientation variants.
Output
Predicted EEG band power per stimulus over time, and the windows where conditions diverge.
Use
Screen out weak stimuli and choose the analysis windows to pre-register before recording participants.
  • Predicted EEG
  • Condition contrast
Participant’s eye-level view of a collaborative robot extending a wooden cube across a table toward themConcept still · generated

04Human–robot interactionBefore HRI studies

Does the motion show people what the robot will do next?

Input
Simulator renders or video of candidate trajectories, approach speeds and handover poses.
Output
Whether predicted gaze reaches the object before release, and arousal as the arm approaches.
Use
Narrow a dozen motion variants to the few worth running with participants.
  • Gaze
  • Arousal
Exploratory

Predictions help narrow down options. Human studies confirm what holds in practice.

Discuss your application
Case study / Advertising

Comparing responses
to two Adidas ads.

This comparison examines predicted attention, arousal, and memory encoding for two Adidas films. The model favored “You Got This,” which also had the higher recorded public view count.

Case study / Two Adidas football adsEqual duration · synchronized playback
Model preferenceB · You Got This

The model favored this film based on its response scores. It also had more recorded public views.

…A views
…B views
Source
A
Football Anthem FilmLoading performance
A / synchronized stimulus
B
You Got ThisLoading performance
Model preference
B / synchronized stimulus
0:00 / 1:28Synchronized playback
Open the full analysis console →

Human response.
New possibilities.

We work with researchers and design teams exploring how models of human response can support their work, from early experiments to applied evaluation.

Request access For research and applied projects