Trending now

The Lanterns midseason trailer, second by second

By the neuro-viral team · 28 Sept 2026 · 4 min read

Lanterns was trending on X in the US on 28 September 2026, so we ran HBO Max's midseason trailer through v2neuro, our brain-encoding model trained on fMRI recordings, to see where the predicted brain response rises and falls across its 1:59.

The peak lands on a line of dialogue

The strongest predicted response in the whole trailer lands at 0:25. The brain system leading at that second is the auditory system, and the words spoken around it are: "There's more than one killer at this table. What?"

That is worth noticing in a trailer. The single strongest moment the model predicts is led by the auditory system, on a spoken line, rather than by the visual system.

Four seconds later, the steepest fall

The steepest fall in predicted response anywhere in the trailer runs from 0:29 to 0:31, only four seconds after the peak.

We can't tell you why from the scan alone, and we won't guess at what is on screen. What the scan does tell an editor is where to look: 0:29 to 0:31 is where the predicted response drops fastest. If this were your cut, that is the stretch you would open first.

A fall like this is not a verdict on the trailer. Every clip has a steepest fall somewhere. The point of knowing where it is, is that an editor can decide whether that stretch is doing the job they want.

A strong first three seconds

For the first three seconds, the predicted response sits above the trailer's own median level. The opening starts high relative to the rest of the clip.

When you scan your own opening, compare it with the rest of your clip in the same way: against its own median, not against someone else's video.

How much to trust this scan

Every neuro-viral scan comes with a per-second trust score, which says how far to rely on the prediction at each moment. On this trailer the mean is 32 out of 100, in the product's fair band: usable, but at the weak end. The scan shows that rather than hiding it.

Read the timestamps above as a prediction of average cortical response. They are not a measurement of any viewer.

What to take into your own edit

Check where your peak is and what carries it. If your strongest predicted moment is a spoken line, give it clean audio and room around it.

Look hard at the stretch after your peak. Here the steepest fall came four seconds later, from 0:29 to 0:31. Scan your own cut and see where yours falls.

Compare cuts rather than trusting one number. Scan the current edit, move or trim the stretch after the peak, and scan it again. neuro-viral's editor does that comparison for you.

Frequently asked questions

What is v2neuro?

v2neuro is neuro-viral's brain-encoding model. It was trained on fMRI recordings of people watching video, and it predicts the average cortical response to a new video, second by second.

Did you use the trailer footage?

No. We scanned the public trailer and show only our predicted response curve. The trailer itself is on HBO Max's YouTube channel.

What does a trust score of 32 mean?

neuro-viral scores every second for how far to rely on the prediction. 32 out of 100 is in the fair band: usable, but at the weak end, so read this scan as indicative rather than precise.

Can I run this on my own video?

Yes. Paste a YouTube link or upload a video at neuro-viral.com. Your first scan is free, for up to 60 seconds of video.

Related reading: where videos lose attention · neuromarketing vs A/B testing.