Before uploading the video, make sure that you instruct the user to record the video that fits the acceptable criteria
– Video Compression
Video collected from user should ALWAYS be compressed. We recommend:
- frame rate: 15 FPS
- dimensions: 320 at longest wide (i.e. whether landscape or portrait)
This is the ffmpeg command line equivalent:
ffmpeg -y -i $INPUT -r 15 -vf “scale=320:trunc(ow/a/2)*2” -strict -2 $OUTPUT
– your users should be instructed to take the selfie image / video at an arm’s length (like a normal selfie)
– should NOT be too close
Currently, there is a maxmium face to frame width ratio that the API will accept.
You can achieve this by placing a “circle” on the screen when your app requests a face video to be taken, like so:
Note: selfie videos taken at arms length often achieves this acceptable face to frame ratio.
– Normal lighting conditions
Not too dark, not too bright. The AI will not be able to “see” with extreme lighting conditions and will result in API to return error.
– OTP / one time password limit
6 digit numeric sequence should be generated and delivered via your application , and passed to the API to determine if the person has spoken it.
– Time Limit
For sake of security, your app should have a time limit between the time OTP is generated and the video is saved.
We live this up to you to implement, but for security sake, we recommend no longer than 20-30 seconds at most since 6 digit OTP can be spoken in less than 5 seconds.