Emotions

The emotion timeline (video)

In short

For videos, Foresight predicts the eight emotions on every frame and draws them as curves in the Temporal trend chart, synced with the player. The overall values average all frames; the Scenes view averages each scene.

What it is

For videos, Foresight predicts the eight emotions on every frame and draws them as curves over time in the Temporal trend chart, under the "Emotions" group. The chart is synced with the video player, so you can see where excitement builds, where a joke lands, where tension peaks and where the emotion fades.

What feeds it

Every frame of the video. The emotion model also uses a short description of the creative's content as context, so it has context about what is shown, not just the single frame.

The emotion analysis runs on your original video, whatever platform you choose.

How it's built

• Curves: one value per emotion per frame, on the same 0–100 scale as the Emotional Profile.
• Overall values: the video's overall emotion values are the average across all frames. Emotional Engagement and Valence for the whole video also summarise the whole timeline.
• Scenes: Foresight splits the video into scenes automatically; the scene breakdown is produced together with the AI insights, so if the insights fail, the Scenes view stays empty until you regenerate them. The Scenes view averages the frame-by-frame values within each scene and shows that scene's Emotional Engagement ("Emo") and Valence ("Val"), next to its attention values. Each frame's Emotional Engagement is positioned against individual frames from the reference videos, and a scene's value averages its frames; the overall score is positioned against whole reference videos. Both use the same 0–100 scale.

The chart supports zooming into a stretch of time and switching individual curves on and off. Attention-side curves (Attention, Clarity, Mental Demand) can be shown too; they are drawn dashed and hidden by default.

Overall video scores summarise the whole timeline.

Scale and direction

Each emotion curve runs from 0 to 100; higher means that emotion is more strongly evoked at that moment. The curves are absolute values with no benchmark. Scene values use the usual scales: Emotional Engagement 0–100, Valence −100 to +100.

Image vs video

The timeline and the Scenes view exist for videos only. An image has a single Emotional Profile.

A video's overall value summarises the average reaction over its whole length, while an image captures one moment. Compare videos with videos and images with images.

How to read it

• Look for the peaks: where does the leading emotion rise, and does it peak at the moment you show the product or brand?
• Watch the arc: a sad opening that turns uplifting shows as falling sadness and rising contentment, awe or excitement.
• Use the Scenes view to find which scene carries the reaction and which ones lose it.
• Overlay the attention curves to check timing, for example whether attention is concentrated when emotion peaks.

Common pitfalls

• Long calm stretches soften the overall values. One strong moment in a long video may barely move them. Check the curve peaks.
• The overall Engagement is not the average of the scenes. Overall and scene values are summarised differently, so they need not match exactly. A video with one intense moment can show a high scene and a moderate overall.
• Curves are lightly smoothed. Frame-to-frame jitter is removed, so a very brief moment can look softer than it was. Read trends across a second or more, not single frames.
• Emotion and attention curves have different meanings. Overlaying them shows timing, not a like-for-like comparison.