The model
Moach
A neural prediction engine. It simulates what a real mind does as it views your advertisement — trained on a large corpus of real MRI, built to model the buyer’s brain.
Most companies cannot say how an ad will perform until they buy airtime. Moach is the engine that answers before the spend. It takes the picture and the soundtrack of your ad — embeddings of sight and sound — and predicts a high-resolution, fMRI-aligned brain response. No scanner session for that cut.
The training is human neuroimaging of people watching video: a large corpus of real MRI. From those hours Moach learned to map vision and audio onto cortical activity. The output is a synthetic BOLD time series across the Schaefer-1000 atlas, one second of predicted brain time per second of the ad.
- What it is
- A neural prediction engine. Video and audio in. A modeled mind out.
- Training
- A large corpus of real MRI — continuous video viewing, human subjects.
- Baseline
- Held-out tests on video Moach has not seen. A shipping baseline, not a ceiling.
- What you get
- A predicted brain response to an advertisement that has never been inside a scanner.
The next iteration is already in training. Version one is for operators who cannot wait until the ads have run to find out whether the work moved anyone.