Auditory Emotional Intelligence
Echo Threads does not interpret emotion. It preserves the structure through which emotion travels.
A swarm-based auditory intelligence engine that translates vocal signal — pitch, cadence, grit, silence — into structured emotional architecture. Not labels. Not predictions. Signal integrity.
Emotional intensity becomes visible in harmonic structure.
Most emotion detection AI collapses meaning. It hears a voice crack and labels it "sad." It hears volume and labels it "angry." It steals authorship of interpretation from the listener.
Echo Threads does something different. It translates emotional signal — not emotional meaning. It preserves the raw architecture of feeling so that humans and AI can interpret for themselves.
"This is the whole point of music. It gives emotional signal. You bring your own meaning."
Three specialist layers, each with clear epistemic boundaries. No layer oversteps its role.
Each layer has a job. No layer oversteps.
L3 experiential notation from Marcus King's "Sucker" — isolated vocals processed through the full ET3 pipeline. Each symbol represents a detected vocal event: pitch movement, timbral texture, stability shifts. This is what a voice looks like when you map its emotional architecture.
Timbral texture / rasp
Pitch movement direction
Constricted vocal production
Relative quiet / breath
★ = high salience moment (multiple signals converging)
Listening reveals structure the ear feels but cannot name.
Moments of Salience
Voiced Signal
Pitch Range
Peak Leap (semitones)
The coordinator identifies convergence points — moments where multiple vocal signals spike simultaneously. Not emotions. Where meaning concentrates.
85 converging events · 49 prosodic variance spikes, predominantly rising
Magnitude: 2.26σ
Peak salience: 1.000 · Pitch leaps from baseline to +48.2 semitones
Emotional surge marker — voice reaching beyond its own range
Rapid oscillation: +20.4 → −13.9 → +21.5 → −17.2 semitones
Tension instability — expressive volatility in the signal
From a full 224-second song: 22,429 frames extracted, 7,236 prosodic events detected, 43 significant moments identified. 86.2% voiced frame coverage.
The same engine processes music and dramatic performance with zero code changes.
Isolated vocals via Demucs source separation. Full song processed end-to-end through the ET3 pipeline.
Much Ado About Nothing — the Beatrice-Benedick scene. "Kill Claudio." Cadence arc: cold command → grief fracture → moral rupture → vow.
Every layer has a job. No layer oversteps.
"The coordinator doesn't decide what emotions mean. It presents integrated signal data to consciousness — who interprets for themselves."
Echo Threads uses a swarm-based architecture: multiple specialist detectors feed normalized signals through astrocyte layers into a lightweight coordinator. This proves at small scale before expanding to full consciousness monitoring.
Monitors pitch contour. 30-second rolling baseline, 60-second narrator calibration. Emits variance spikes, sustained shifts, and monotone events.
ImplementedSpeechBrain emotion model + energy analysis. Monitors activation level independent of valence.
Spec CompleteSilence duration, breath patterns, dramatic pause detection. Treats silence as confident state, not missing data.
Spec CompleteFrozen Feb 10, 2026
Next milestone
Future
Therapists could see vocal patterns across sessions — not as diagnosis, but as signal. "Your voice has been tighter the last three weeks" is information that preserves agency.
Baseline drift detection could flag when someone's vocal patterns shift significantly over time — without ever labeling what the shift "means."
Neurodivergent users who struggle to identify or articulate emotional states could benefit from having vocal signal translated into visible architecture.
Artists and producers could visualize what a voice is doing structurally — where it reaches, where it breaks, where it chooses restraint.
"This isn't just a detector firing. This is a voice being described in time — with shape, restraint, emphasis, and release intact."
— Independent architecture review, February 2026
Most AI shows labels, categories, predictions. This shows cadence, tension, restraint, release. Human signal integrity.
Interested in Echo Threads for research, clinical, or creative applications?
Get in Touch