How videos are analysed
Foresight predicts emotion on every frame and attention on frames sampled throughout the video, giving continuous curves, scene scores and overall scores that summarise the whole video. Video scores are positioned against reference videos and are not interchangeable with image scores.
A video is analysed moment by moment. You get continuous curves showing how attention and emotion evolve, scores for each scene, and overall scores that summarise the whole video.
What feeds it
Videos can be up to 90 seconds long. Every video is first converted to a standard working format (a standard frame rate and working resolution) so that all videos are processed the same way. Animated GIFs are treated as videos.
- Emotion is predicted on every frame of your original video. The emotion model also uses a short description of the creative's content as context.
- Attention: Foresight analyses sampled frames across the whole video, and the moments in between are filled in. This follows cuts and changes in the picture over time while keeping processing time reasonable. If you chose a platform, every frame is analysed inside that platform's screen.
Sound is not analysed directly; music and sound effects do not affect the scores. Spoken words can be reflected indirectly when the content description mentions them.
What it produces
- Overall scores: the same KPIs as for images: E-Score, Emotional Engagement, Valence, the eight emotions, Attention, Clarity, Focus, Spread and Mental Demand.
- The Temporal trend: curves for the eight emotions, Attention, Clarity and Mental Demand over time, on the Metrics tab. See The emotion timeline.
- Scenes: Foresight splits the video into scenes automatically. Each scene shows the average of its moment-by-moment values and a colour: Good, Medium or Low, following the same bands as the E-Score.
- Heatmap and fogmap videos you can play alongside the original.
- AI insights, including a scene-by-scene analysis and recommendations you can jump to with Play scene.
The gaze plot, attention build-up and areas of interest are available for images only.
How the overall scores are built
The overall scores summarise the whole video, giving every moment the same weight. Foresight combines the moment-by-moment predictions over the full timeline into one overall result for each KPI:
- the overall emotional profile is the average profile across all frames;
- the overall attention scores summarise the attention predictions over the whole video.
Moment-by-moment and scene values sit on the same 0–100 scale as the overall scores, so you can compare a moment with the video's overall result.
Benchmarks for videos
Videos have no creative-type or sector benchmarks, and no comparison value is shown for them, with or without a platform. Your scores are still positioned against the reference library of videos, and timeline values against individual frames from those reference videos. See Platforms and benchmarks.
Image vs video
Video and image scores share KPIs, scale and colour bands, but they are not interchangeable. Every score is positioned against creatives of the same media type (see how scores are positioned), the references are different, and a video score summarises a whole timeline while an image score describes a single view. Compare videos with videos, and use the colour bands if you need a rough read of a video next to an image. See Image vs video benchmarks.