Loading product…
04Photo archive with face recognitionPrototype
MediaVault
MediaVault keeps an organisation's photos private and recognises who is in each one. Upload a batch and every enrolled person is tagged automatically, so gathering every photo of someone stops taking an afternoon. Galleries can be public, password-protected or private.
01Problem
The problem it solves
An organisation that photographs the same people again and again ends up with thousands of pictures and no quick way to answer a simple question: which photos is this person in? Tagging faces by hand takes hours, and it never keeps up.
MediaVault does the tagging on the way in. It learns each person from a reference photo, then finds them in every new photo that's uploaded, and keeps the archive private while it does.
02How it works
Step by step
- 01
Drop
A photographer drags a batch of photos into Upload and adds tags, such as the occasion, in the same step.
- 02
Compress
Each photo is compressed in the browser before it leaves the computer, so big batches upload quickly.
- 03
Store
Photos go into private storage, one folder per organisation, through links that are checked against the organisation and expire after an hour.
- 04
Scan
A queue hands each photo to face detection, which finds and crops every face in it, so group photos work too.
- 05
Match
Each face is compared with the people the organisation has enrolled. A match at 90% or better links that person to the photo.
- 06
File
Admins build galleries from tags or from people, and decide who can open each one.
03Drawing
Drawn as a machine
A vault line: photos are compressed, signed in, scanned for faces, matched and filed into galleries. Each part of the drawing is a real part of the product. Hover or tap one to see what it does.
Part 06 / 08
Face scanner
Finds and crops every face.
36 faces found
- Uploaded
- 021
- MB in → out
- 72 → 59
- In queue
- 02
- Faces
- 036
- Matched
- 020
- Last match
- 93.8%
- Link TTL
- 3599 s
- Filed
- 09·04·02
04Features
Everything it does
Upload
Get a whole shoot in at once.
- Bulk drag and drop
- Drop many photos at once; JPG and PNG.
- Tags on the way in
- Pick existing tags or create new ones while uploading.
- Upload log
- Every photo with who uploaded it and when, and a full-screen viewer.
Recognition
Who is in each photo, without tagging by hand.
- People from reference photos
- Add a person with at least one reference photo; each one is checked to contain exactly one face.
- Group photos
- Every face in a photo is matched on its own, so one photo can link to several people.
- A clear status
- Each photo is marked done or failed, with a plain reason such as "image is too big".
- 90% threshold
- A face only counts as a match at 90% similarity or more.
Galleries
Collections that build themselves.
- From tags or people
- A gallery collects every photo with a tag, or every photo of chosen people.
- Three access levels
- Public, password-protected, or private to the users you pick.
- Viewing
- Photos grouped by date, newest first, with a full-screen viewer.
Organisations
One archive per organisation, kept apart.
- Invitations
- Invite people by email with a role; invitations show as pending, accepted or expired.
- Three roles
- Admins run everything, photographers only upload, users see the galleries they're given.
- Several organisations
- Switch between organisations, each with its own name and logo.
- Five languages
- English, Spanish, French, Portuguese and German.
05Fit
Who it's for
Who it’s for
Organisations that photograph the same group of people again and again, and need to find everyone fast.
Where it stands
PrototypeA working prototype built in 2023. Not publicly launched.
06Specification
The numbers
| Formats | JPG · JPEG · PNG |
|---|---|
| Browser compression | Quality 0.8 |
| Face match threshold | 90% |
| Max image for detection | 5 MB |
| Signed link lifetime | 3,600 s |
| Gallery access | Public · password · private |
| Interface languages | 5 |
| Products used to build it |
|
07Questions
Straight answers
01Can I use it?
Not yet. MediaVault is a working prototype from 2023 and hasn't launched. If it would fit your organisation, ask for a demo.
02How does it know who someone is?
An admin adds the person with at least one reference photo. Each reference is checked to contain exactly one face before it's used for matching.
03Who can see the photos?
Files stay private in storage and open only through links that expire after an hour, checked against the organisation. Photographers can only upload; users only see the galleries they've been given.
04What isn't finished?
Search, downloads and sharing from a gallery, galleries that mix tags and people, and the password check on password-protected galleries. Upload, recognition and the gallery basics work.