Here are answers to some of the most common questions we receive.
WHAT IS THE ACCURACY OF THE VOICE BIOMETRICS?
VoiceIt is proud to provide customers, after Model Tuning, with a FAR of 0.0001% complemented by a FRR of 2% - 10%.
WHAT IS MODEL TUNING?
This process involves taking real-world data to fine-tune the Voice Engines and make sure they are optimized for a customer's specific channel, codec, voiceprint phrase, etc... Best of all, it's included for every developer!
WHAT IS THE ACCURACY OF THE FACIAL BIOMETRICS?
Against the popular LFW Database, VoiceIt scores an impressive 99.38%. We are currently conducting testing with NIST and hope to be able to provide further results soon.
WHAT IF SOMEONE RECORDS MY VOICE?
Our biometrics engines automatically reject any authentication that matches too closely with a previous recording, as there will always be subtle differences each time you speak.
CAN I AUTHENTICATE IN ANY LANGUAGE?
We support authentications in many languages. Text dependent characteristics and text confidence detection are currently supported in 88 languages
WHY DO I HAVE TO SAY AN APPROVED VOICEPRINT PHRASE?
Contrary to popular belief, "My voice is my password," is an unsecured phrase. Due to repeating IPA phonetics (sounds) and length, it does not meet our security standards. Approved phrases contain enough data and IPA phonetics to allow for an accurate voiceprint. Phrases can be requested and managed under the Voiceprint Phrases section of the developer console. Requested phrases should have no repeating IPA phonetic and be a minimum of 8 syllables.
WHAT MAKES A GOOD VOICEPRINT RECORDING?
Speak in a normal tone and speed. Please make sure to enroll and authenticate 1 full second after the beginning of the recording. We recommend using a cue or trigger to help your users know that the recording has started. Watch Video
WHAT IF THERE IS A LOT OF BACKGROUND NOISE?
VoiceIt incorporates noise-cancelling and voice detection technology for enrollment and authentication.
WHAT IF I HAVE A COLD?
Our biometrics engines intelligently enrolls certain successful authentications and tracks gradual changes in your voice over time. For some sudden changes, such as sickness, your voice may be too different to authenticate. However, this only happens in extreme cases.
MY RECORDING WAS UNSUCCESSFUL. WHY?
Here are some possible reasons that your recording might have failed: Your recording was too short or too long (must be at least 1.3 seconds long and no longer than 5 seconds). You did not use an approved voiceprint phrase. The voiceprint phrase spoken was too short. Background noise may have been too loud. Recording volume may be too low.
WHAT AUDIO CODEC ARE SUPPORTED?
The recommended codec is PCM (wav RIFF) with 44.1KHz to 8KHz sample rate, 1 or 2 Channels, 48 to 8 Bits. However, the following codecs are supported: 8svx aac alac aif aifc aiff aiffc al amb au avr cdda cdr cvs cvsd cvu dat dvms f32 f4 f64 f8 fssd gsm gsrt hcom htk ima ircam la lpc lpc10 lu maud mp2 mp3 nist prc raw s1 s16 s2 s24 s3 s32 s4 s8 sb sf sl sln smp snd sndr sndt sou sox sph sw txw u1 u16 u2 u24 u3 u32 u4 u8 ub ul uw vms voc vox wav wavpcm wve xa.