00:00:00/Video intelligence
Point a camera at anything.
Ask it in plain English.
Svid watches your footage (CCTV, an upload, a URL, or a live phone) and turns every frame into words you can search. Ask a question, get the moment. Add a check, know the instant it happens.
No tracking, no facial recognition · pay as you go from a $5 top-up
5 minutes or 30: tell us what you want your cameras to catch and we'll work out how to do it. Pick a time.
Get the Svid app on iPhone and Android
Svid reads the frame
Built for the places cameras already watch
00:01:00/The pipeline
Video in. Answers out.
Svid is an interface, not a single-purpose app: video becomes frames, frames become descriptions, descriptions become search and alerts. The core never changes, only the questions you ask of it.
- Ingest01
Point it at any footage
An RTSP camera, an uploaded file, a video URL, or a live phone. Live and recorded are the same to Svid: it only ever needs an image out of the stream.
- Read02
It watches at your level
Frames are sampled at your chosen frame rate (from 1 every 15 seconds up to every frame), skipped when nothing moves, and described in plain English by a vision model.
- Search03
Ask in plain English
Every moment becomes searchable. Type "person in a red jacket near the till" and jump straight to it. No scrubbing, no timestamps.
- Check04
Checks alert and report
A check judges a window of recent frames, so a fall or a concealment reads as an event, not a snapshot. Alerts land on screen, by email, or by webhook, and reports arrive written.
00:02:00/The library
Every camera and clip, one searchable library.
Footage lands in a Finder-style library: folders inside folders, live feeds next to recordings, drag-and-drop to reorganise. The search bar reads the folder you're standing in, subfolders included.
It searches meaning, not keywords, so "someone struggling with heavy boxes" finds the moment even though nobody ever typed those words. A check can be run across a whole folder in one action.
stores / london / entrancesearches subfolders too
- 01"person in a red jacket near the till"00:42
- 02"forklift reversing without a spotter"02:59
- 03"someone slipped in the entrance"05:16
- 04"child left the play area alone"07:33
- 05"queue longer than four people"09:50
00:05:00/Live & delivery
From live feed to your inbox in seconds.
Open a phone's camera, or point Svid at the RTSP feed from a recorder you already own, and it becomes a live sensor: each frame judged by every check you've turned on. When something matters, you hear about it where you already work.
Weapon alert
Flagged 14:22:07 · banner, buzz & email sent
One event, one alert
Everything flagged lands in the Alerts tab, filterable. A check re-firing within 2 minutes is treated as the same event continuing, not a new siren.
Alerts and written reports can land in your inbox, so nobody has to sit watching a dashboard.
Webhooks
A JSON payload on every alert wires Svid into your POS, Slack, or any system you already run.
Every live session is saved as a video too, so live footage is searchable the moment it's captured. No new hardware: a signed-in phone or the cameras already on the wall.
00:06:00/Pricing
Pay as you go, no surprises.
No plans, no tiers. Top up from $5 and spend it as frames are analysed (bigger top-ups earn up to 20% bonus credits). Every account starts with $5 of free credit, $10 more when you add a card, and every check is free.
Pay as you go
Top up from $5
Unlimited cameras · every resolution
One way to pay, and it's the fair one: top up from $5 and only pay for the credits you actually use, shown in dollars as you go.
- ✓Every camera, live and RTSP, no limit
- ✓Every resolution & frame rate
- ✓Upload & search recorded footage
- ✓All alerts & reports
- ✓$5 of free credit the moment you sign up (no card needed)
- ✓$10 more when you add a card
- ✓Every check is free: you only pay for analysis
- ✓Auto-refill on by default: you set the amount and the threshold, and switch it off any time
- ✓5% extra credit on every auto-refill
What will it cost?
The same 2 dials the app asks for. Every frame gets a written description: more words catch smaller details and cost more, and more frames a second catch faster events and cost more.
1 · Words per frame
2 · Frame rate
- ≈ $0.554 to $0.697
- an hour of video
- ≈ $13.30 to $16.72
- 24 hours of video
3,600 frames an hour, one analysis each. Resolution doesn't change the price and checks are free. That's the ceiling, with every sampled frame analysed. Motion gating usually skips about 40% of still frames, so an hour typically comes to ≈ $0.332 to $0.418.
$112bn
lost by US retailers to shrink in a single year, most of it theft that cameras recorded and nobody watched.
National Retail Federation, 2023
$20,000+
where the average slip-and-fall claim settles. Timestamped footage of the hazard, or its absence, decides most of them.
National Floor Safety Institute
£20.7bn
what workplace injury and ill health cost UK employers in a year.
UK Health & Safety Executive, 2022/23
1:1
reviewing footage manually takes as long as the footage. An hour of video is an hour of someone's day. Search answers in seconds.
Arithmetic
Sign up ($5 of free credit, $10 more when you add a card) and process your first footage in minutes.
00:07:00/FAQ
Questions, answered.
Do you track people or use facial recognition?
No, deliberately. There's no facial recognition, no biometrics, and no tracking of people anywhere in Svid. The only AI in the product describes images, checks conditions on them, and embeds the descriptions for search.
How is the price worked out?
Pure pay-as-you-go: top up from $5 and only pay for the credits you actually use, shown in dollars as you go. Bigger top-ups earn bonus credits, up to 20% at $1,000. Auto-refill is on by default so cameras never stop mid-event ($10 when you run low), and you can switch it off any time. Every account starts with $5 of free credit (no card needed), and adding a card earns $10 more. Checks are free: you only pay for the analysis they run.
Does it work with our existing cameras?
Yes. If a camera produces footage, Svid can read it: an RTSP feed, exported files, a video URL, or a phone propped on a shelf. There's no special hardware.
Is it live or after-the-fact?
Both. Recorded footage becomes searchable, and a phone or IP camera can be monitored live, raising alerts as frames come in. Live sessions are saved as videos too, so they're searchable afterwards.
Where does the footage live?
In your account and nowhere else, and it's never used beyond it. Processing sends frames to the vision model to be described. There isn't a fully on-prem mode yet. If you need fully on-prem, privacy-oriented search, leave a message in the chat at the bottom right and we'll get back to you.