Engagement analysis for video and audio
see what they feel.
fix what loses them.
Upload a clip. In about a minute, NeuroVidz scores its hook, its hold and its feel from what is actually in the picture and the sound — second by second — and lists the edits to make, each pinned to a timestamp.
- 39 signals, every second
- Same file, same score
- Video or audio, up to three minutes a read
- Free to start
Second by second
Craft score
It keeps moving and has real peaks, but the opening is the weak spot: first word at 4.8 s.
Fix first
- 1
Start talking sooner — the first word lands at 4.8 s. Trim the lead-in so it lands under 1 s.
Hook+5.6 to the score - 2
Front-load the energy — the opening is calmer than the rest. Tease the peak at 0:08 in the first seconds.
Hook+5.6 to the score
Made for the feed
Every clip fights the scroll.
Reels, Shorts and TikToks, ads, vlogs and music videos — and podcasts, audio only. Up to three minutes a read.
Illustrative clips and example numbers — not real reads. Hover or tap a clip.
How it works
Three steps, about a minute.
- 1
Drop a clip
Video or audio — mp4, mov, webm, mp3, wav. Anything longer than three minutes, you pick the section to read.
launch-teaser.mp40:22Reading picture and sound…
- 2
Every second is measured
Cuts, motion, faces and expressions; speech, pace, pitch, music, beat and loudness. Each measurement is tested against clips with a known answer.
PictureSpeechIntensity39 signals, every second
- 3
Get the score and the edits
Hook, Hold, Feel and Story, every check behind them, and the fixes ranked in the order a viewer meets the problems.
Fix first
Start talking sooner — the first word lands at 4.8 s. Trim the lead-in so it lands under 1 s.
▸ 0:05Hook
Hook
The first three seconds, checked one by one.
Does something happen at once? Does the first word land inside a second? Is there a face, movement, a sound? Is the opening as lively as the rest of the clip? Each check says what it measured — “first word at 4.8 s” — and what to do about it. Checks that cannot apply to a clip say so instead of counting against it.
Hook
43
Something happens from the first second
starts at once
The picture changes in the opening
a small change in the first 3 s
The first word comes fast
first word at 4.8 s
A face appears early
first face at 0:03
The opening is as lively as the rest
opening at 56% of the clip's typical intensity
The opening is loud enough
-15 LUFS in the first 3 s
Hold
Every second on one line — scrub it.
Intensity, picture change, speech and music, and the moments that pull attention back: a cut, a voice starting, a face arriving, a jump in loudness. Dead air and long still stretches are flagged where they happen. In the app the line is the player's scrubber; here, click anywhere on it.
A 22-second talking clip — click anywhere
Feel · sound
It listens as closely as it watches.
When the first word lands, how fast the talking is, how much the voice moves, where the music sits and how loud the mix plays after a platform evens it out. Podcasts and voice notes get the full read too — audio only, no picture needed.
Numbers from the same 22-second example. Loudness follows the broadcast standard (ITU-R BS.1770), the way YouTube and Spotify measure it.
First word
4.8 s
when the talking starts
Speaking rate
5.7 /s
syllables per second of speech
Voice movement
3.8 st
how far the pitch moves, in semitones
Speech
74%
of the clip
Loudness
-14 LUFS
measured the way platforms do
Loudness range
3 LU
between the quiet and loud parts
Story
And a read of the story itself.
A language model reads what is said and shown and answers seven plain questions, each with the moment it is judging. It counts as one part of four and is always marked as a model's read — the other three parts are measured from the file.
- 1Promise
Do the first 3 seconds say what the viewer gets by staying?
- 2Payoff
Is that promise delivered?
- 3Clarity
Is the point easy to follow on one watch?
- 4Specific
A real example, number or demo — not a general claim?
- 5Surprise
At least one moment that breaks expectation?
- 6Stakes
A human reason to care?
- 7Ending
Does it land, or trail off?
Who it is for
Built for the edit before you post.
Short-form creators
Reels, Shorts and TikTok: fix the first three seconds and the dead spots before you post.
Checks: first word and first face · still stretches and dead air · where the high point lands
Podcasters and voice
Audio-only reads: long pauses, pace, how much your voice moves, and where it drags.
Checks: silences over a second · syllables per second · loudness against platform level
Brands and agencies
Compare two cuts of an ad before the budget goes behind one — same checks, side by side.
Checks: hook of each cut · new moments per 10 s · the ending
Educators and explainers
Long talking heads: find where attention sags and what to cut or cover with b-roll.
Checks: sagging stretches · picture change · vocal expressiveness
Musicians and music videos
Intro, build and drop: where the energy rises, where it stalls, and how the video carries it.
Checks: tempo and beat · loudness range · cuts against the music
Measured, tested, plain about its limits
Why the numbers hold up.
Tested against known answers
Every measurement is checked against clips built so the right answer is known. The checks are part of the product's test suite.
- A metronome at 120 BPM
- reads 120 BPM
- Major and minor chord progressions
- read major and minor
- A test tone, white noise, chords
- read 0% speech
- A camera panning across a still scene
- reads as a camera move, not a moving subject
- The same voice 6 dB quieter
- reads the same pace, pitch and speech share
- 2 seconds of dead air added to a real clip
- lowers its Hook by 22–62 points
Same file, same score
The three measured parts are deterministic. The story read runs at a fixed setting and is one part in four, so it cannot swing the score on its own.
Not a brain scan
The 3D brain in a result is a model of the measured signals — a way to see them, not a recording of anyone's brain.
Missing is never zero
No faces in a clip, no speech, or expression reading switched off: those checks are left out and named, never scored as a failure.
Questions
Asked, answered.
What is NeuroVidz?
A tool that reads a video or audio clip second by second — what is in the picture and what is in the sound — and scores how well it hooks, holds and moves a viewer, with the exact edits to make, each pinned to a timestamp.
How does it work?
It measures 39 signals every second — cuts, motion, faces and expressions, speech, pace, pitch, music, beat and loudness — and checks them against what keeps people watching: a fast start, no dead air, something new every few seconds, a clear high point. A language model then answers seven plain questions about the story. Every part of the score shows what it measured.
How is the score calculated?
Four parts — Hook, Hold, Feel and Story — each the average of its checks, and the score is the average of the parts. Checks that cannot apply to your clip (no faces, no speech) are left out and named, never counted against it. The score tells you whether a clip does the things that hold attention; it does not promise views.
Is this a real brain scan or medically accurate?
No. Nothing is measured from anyone's brain. The score comes from the clip itself. The 3D brain in a result is a model of the measured signals, shown as a way to see them. None of it is medical, diagnostic or clinical.
Do you recognize people or store face data?
No. We detect faces only to read expression signals — we don't identify anyone or build facial-recognition templates, and face crops are discarded after processing. You can switch expression reading off in settings.
What happens to my video and data?
Your video is processed to produce the analysis and stored securely in your account. You stay in control — export or delete everything whenever you want.
What can I upload, and how long does it take?
Video or audio up to 400 MB (mp4, mov, webm, mp3, wav). Each read covers up to three minutes; drop a longer file and you pick which three minutes to read, right in the uploader. The measured score lands in about 30 seconds and the story read about a minute later.
Can I export or delete my data?
Yes — anytime, from your account. Export downloads a copy of your data; delete permanently removes your videos and analyses.
Try it on your next clip.
Free to start. The score and the edits land in about a minute.
The example on this page is a real result: a 22-second talking clip, scored 76.