What it looks like in practice
The app shows questions like «what does an evening at home mean?», «what's the favorite joke?». Recording is started — front camera, up to 60 seconds per answer. It is possible to skip, re-record, change. All answers stitch into the mmeet-video.
The recording isn't immediately public. It's served via signed temporary URLs to authenticated clients — after the person is seen in the discovery feed and matched with. The video isn't sitting in an open feed, isn't indexed by search engines.
Why questions, not free-form
Just say «tell me about yourself on camera» and most people freeze. Without structure there is either a rehearsed monologue or awkward silence. Questions remove that pressure. Instead of «I have to come up with what to say about myself», specific questions are answered like «what's your favorite season?». It's a mini-interview, not a performance.
Why one video, not a new one per match
Time. Recording a new video for each match means stepping in front of the camera every time someone likes the profile. Doesn't scale.
Authenticity. If it is known a specific person will see this video, calibration starts to please them. One video for «everyone in general» is a more honest self-portrait.
Re-recording is possible anytime — old gets deleted, new replaces.
What about privacy
The mmeet-video isn't shared with third parties. Stored in our S3 with ACL=PRIVATE, served via signed temporary URLs that expire. If an account is deleted — the video is removed from our storage. Devices that already loaded the video earlier may keep a short-lived player cache until refresh — that's outside our control.
We don't use videos to train models, don't pass them to ad networks, don't analyze content for targeting.