Aevlys AI · Voice DNA
Voice DNA: Decode Your Speaking Style.
Speak a 60-second pitch. Voice DNA measures pace, fillers, pauses, pitch and your ending against fixed standards, then tells you what failed, when it failed and how to fix it.
PREVIEW · THE HELIX MOVES WITH YOUR VOICE IN THE ARENA
Practice Studio
Timing lightsPick a format to rehearse. Each format runs the green, yellow and red lights, counts your filler words, and ends with a Strength, Next step, Strength review.
What gets decoded
- Pace
- Average and 10-second peaks. The pitch standard is 130 to 160 WPM. Spikes are timestamped.
- Fillers
- Rate, clusters and dead air. um, uh, like, you know and more, per 100 words.
- Voice
- Pitch variation and ending volume from the desktop microphone. Flat delivery is flagged.
- Ending
- Cut off, trailing off or no close. Every flaw comes with a timestamp and a drill.
Recent decodes
Saved on this deviceFailure patterns
Named after your worst measured faultThe Arena
Coach strictness
Your 60-second elevator pitch
1:00
remaining
Press start when you are ready.
1:00
1:30
2:00
Live transcript
Fillers glow amberWords0
Pace0
Fillers0
Your words appear here as you speak.
Type or paste your pitch
Typed modePace and timing come from the delivery time you enter. Energy is estimated from the text.
Word challenge
Used 0 times so far. Work it in naturally.
Verdict
0/100
Grade
Re-score this take as
Your voice signature
Critical flaws
Corrective drills
What worked
Speaking pace
Ideal
0WPM
Energy
Strong0%
Clarity index
Strong0/100
Session report
Inside windowTiming report
0:00
Filler report
Word challenge
Filler word detector
Pattern matched on your transcriptDetected
0
Per 100 words
0.0
Standards in force
Words0
Duration0:00
Power words0
Hedges0
How these scores are calculated
- Flags, not averages. Each metric is checked against a fixed standard and rated Watch, Red flag or Critical. A flag costs 5, 12 or 20 points, scaled by the coach mode (×0.6, ×1, ×1.5), with a ceiling per category so one problem cannot sink the score twice.
- Caps. Any critical flag, two red flags or one red flag cap the score at a mode-specific ceiling. Anything above 89 needs zero watch items and every metric measured. A score of 0 is reserved for no speech.
- Timestamps come from when the browser delivered each phrase, so they are accurate to roughly one to three seconds. Typed takes have no timestamps and show word positions instead.
- Pitch, loudness and pauses need the desktop microphone meter. When they cannot be measured, the report says so and the score is capped at 89. Nothing is estimated from text.
- Fillers are matched by pattern. Browser speech recognition often leaves out “um” and “uh”, so voice counts are a floor. Recognition confidence is not used in any score because it penalises accents and poor microphones.
- Thresholds are starting values for a prototype and should be calibrated on real recordings. The critique text is generated by rules from the flags. It is not a validated assessment.