Goal profiles
A goal profile decides which reference bands the dashboard overlays on your primary cues. Profiles affect interpretation only — they do not change which acoustic measures are computed or how they are computed. Switching profiles re-renders an existing result; it never re-measures.
The four built-in profiles
| Key | Display name | Target band on primary cues |
|---|---|---|
feminine_reference |
Feminine reference | Feminine |
masculine_reference |
Masculine reference | Masculine |
androgynous_reference |
Androgynous reference | Androgynous |
custom_tracking_only |
Custom / tracking only | No classification rendered |
The three "reference" profiles share the same underlying band data
(from src/vocal_analyzer/models/ranges.py) — they just differ in
which band is labelled as the target. The custom_tracking_only
profile skips classification entirely, surfacing the raw measurement
value without the band-colour cue.
When to pick tracking-only
Pick custom_tracking_only when:
- You want the measurements without the classification framing.
- You are training toward a target that does not map cleanly onto "masculine", "feminine", or "androgynous" — for example, a stylistic goal based on a specific speaker you want to sound like.
- You are using Vocal Analyzer primarily to compare against yourself over time, not against population bands.
The tile values and trends stay available; only the coloured band overlays and the percentage-of-band positioning go away.
What stays the same across profiles
- Every raw measurement (F0 median, formant medians, dispersion, HNR, voice quality, etc.). These are computed from audio, not from profile choice.
- Confidence badges and measurement warnings.
- Transcription, vowel-space charts, timeline canvas.
- Session history.
What changes across profiles
- The diamond position and band colouring on the spectrum bars. A pitch median in the feminine band looks different against the feminine profile (in-target) than against the masculine profile (out-of-target).
- The percentile math for coherence. Coherence compares F0 and tilt percentile positions inside the active profile's bands, so a switch of profile can move the classification from Matched to Divergent even though the raw measurement is unchanged.
- The tile labels that reference target bands (for example, the resonance tile's "Masculine / Androgynous / Feminine" classification text).
Changing the active profile
Today, the profile is set on the Settings page (in the installed
app, also through the CLI config file). The dashboard does not have a
live profile-switcher UI in v1 — changing profile requires a rerun of the analysis with
the new profile selected. See
Configuration for where the goal_profile
setting lives.
Why "descriptive overlays, not clinical cutoffs"
Profiles are a convenience layer on top of band data that is itself descriptive:
- Band boundaries come from published central tendencies, not clinically validated perceptual cutoffs.
- Two speakers with identical F0 medians can land very differently with real listeners.
- Self-perception and listener perception are independent dimensions (Dacakis et al. 2017); no profile choice resolves that dimension.
If you are using Vocal Analyzer alongside active voice training, a voice teacher can translate the dashboard numbers into training targets in a way no profile overlay can.