This is a sample report for a 24-second talking-head video, so you can see everything a scan gives you.
- Attention Control81
- Visual Cortex77
- Auditory Cortex69
- Language Network58
- Emotional Response63
- Focus Drift34risk
Attention map inspired by neuroscience research — not a brain scan.
Sample report · morning-routine-v2.mp4 · 24.0s
Good — a few tweaks away
Strong open, soft middle
Your first second grabs attention with movement and your face on screen. Viewers are most likely to drift between 9s and 14s, where the shot holds still and nothing new appears.
Hook score
0
first 3 seconds
Hold estimate
0%
watch to the end (est.)
Drift risk
0
higher is worse
A prediction to guide your edit, not a guarantee of views.
5 fixes · most impact first
What to fix before you post
Break up the still stretch at 9–14s
High impactAdd a punch-in zoom, a B-roll cutaway or a text callout around 9s. Nothing changes on screen for almost 5 seconds.
PacingAdd a text hook in the first second
High impactPut the promise of the video on screen in big text — e.g. “The 3-second rule nobody follows”. Many people decide with the sound off.
HookCaption the whole video
Medium impactCaptions only cover 18% of the runtime. Burned-in captions keep muted viewers watching and make every word land.
Captions & speechTighten the two silent pauses
Medium impactCut the gaps at 10.4s and 17.8s, or fill them with a beat. Silence is where thumbs start moving.
AudioEnd on the payoff, then loop
PolishThe last 2 seconds lose energy. Finish on the reveal and cut so the ending flows back into the first frame.
Pacing
Already working
- Strong opening — motion, a face, fast speech in the first seconds
- Good pacing: a new shot every 4.8s
- Clear, well-levelled audio
Exact time · frame number · what to change
Frame-by-frame edit plan
9 timestamped edits · 30 fps · 720 frames
Sample report — scan your own video to step through it frame by frame.
0:00.00 · f0
The mechanical edits, done for you · on your device
One-click fix
24.0s → 22.6s · 1.4s tighter
Sample report — scan your own video to make a fixed version.
First 3 seconds · AI editor
Your hook
You, mid-frame in a bright kitchen, reaching for a mug — clear subject and natural motion.
“Nobody tells you this about mornings.”
Good curiosity gap and a face in frame 0, but the line arrives after a breath and nothing on screen tells muted viewers what the video promises.
Try this text hook
“The 3-second rule nobody follows”
Try this opening line
“This 3-second rule fixed my mornings — and nobody talks about it.”
Every signal we measured
Deep report
Retention estimate
44% still watching at the end · bars show attention pull per second
AI editor's judgement
From the frames and transcript, compared with strong short-form in your niche (50 = typical).
- HookStopping power of the first 2 seconds
- 74
- ClarityInstantly clear what it's about
- 82
- ValuePayoff for the viewer's time
- 71
- EmotionThe feeling it triggers
- 58
- PacingKeeps moving, no drag
- 54
- PayoffSatisfying end or reason to rewatch
- 66
- ShareabilityWould someone send it to a friend
- 61
Attention Control
81- Motion in first 3s
- LivelyGood
- Face on screen
- 0.2sGood
- Text hook
- None foundWeak
- First words
- 0.6sGood
Visual Cortex
77- Cuts per 10s
- 2.1Good
- Average shot
- 4.8sGood
- Motion
- LivelyGood
- Contrast
- 31%Good
- Brightness
- 46%Good
- Sharpness
- SharpGood
Auditory Cortex
69- Loudness
- -19 dBFSGood
- Silence
- 14%OK
- Opening energy
- +1 dB vs avgGood
- Speech · music
- 71% · 8%Good
Language Network
58- Captions on screen
- 18%Weak
- Speaking pace
- 2.8 words/sGood
Emotional Response
63- Close-up time
- 42%OK
- Expression changes
- 3OK
- Audio peaks per 10s
- 1.3Good
Focus Drift
34 risk- Longest static shot
- 4.6sWeak
- Dead air
- 1.2sOK
- Silent gaps
- 2OK
What we heard
- Nobody tells you this about mornings.
- Um, there's a three-second rule, and once you see it you can't unsee it.
- The second your alarm goes off, you count down from three and you stand up. No snooze, no scrolling.
- I've done it for thirty days and this is the difference.
Measured with: faces opencv-haar · text heuristic · speech whisper · writer llm · frames_reviewed 36 · signals v1.0-rules